Bioengineering Strategies to Address Key Bottlenecks in Ferroptosis-Based Cancer Therapy: A Critical Review

Abstract

More than a decade after ferroptosis was first defined, no agent designed to exploit it has won clinical approval for cancer. We argue that the bottleneck has shifted from mechanism to translation: unsatisfactory pharmacokinetics with unreliable intratumoral accumulation; resistance arising from single-pathway blockade; a scarcity of selective, well-tolerated modulators; and the absence of non-invasive tools for monitoring target engagement. Instead of cataloging nanoplatforms by material class, this critical review organizes bioengineering strategies into four levels of rising functional integration, namely node-specific delivery, multi-pathway combination, stimuli-responsive gating, and theranostic integration, each mapped to one distinct bottleneck. The ordering is ordinal, not evaluative: the most clinically advanced ferroptosis nanomedicine, carbon nanoparticle-iron(II) complex (CNSI-Fe(II)), is a Level 1 platform. Literature was retrieved through systematic searches of PubMed and Web of Science up to June 2026. Synthesizing available preclinical and preliminary clinical findings, this work identifies three cross-cutting translational issues that warrant critical reassessment. Only one ferroptosis nanomedicine has reached human testing; nearly all supporting evidence derives from subcutaneous xenografts in immunodeficient mice, with safety follow-up rarely extending beyond acute timeframes. Passive tumor uptake relying on the enhanced permeability and retention (EPR) effect yields highly variable tumor accumulation that cannot be reliably forecasted for human solid tumors. Non-degradable metallic nanozymes bring unaddressed long-term organ risks originating from systemic metal deposition. Combination treatments administered without companion diagnostic tools fail to filter patient subgroups with treatment sensitivity. We contend that progress now depends less on more elaborate nanocarrier architectures than on predictive large-animal toxicology, ¹⁸F-FSPG PET-guided patient stratification ((4S)-4-(3-[¹⁸F]fluoropropyl)-L-glutamate positron emission tomography), and locoregional delivery that limits systemic exposure.

Introduction

Ferroptosis, an iron-driven form of regulated cell death driven by lipid peroxidation, exhibits therapeutic potential across diverse human disorders including cancers, neurodegenerative diseases and ischemia-reperfusion injuries. Still, no therapeutic agent targeting ferroptosis has obtained formal clinical approval.^1–3 In oncology, induction of ferroptosis has been shown to reverse chemoresistance in multiple cancer types and suppress metastatic progression.⁴ In neurodegenerative conditions such as Parkinson’s disease and epilepsy, inhibiting ferroptosis in neuronal cells alleviates motor dysfunction and reduces parenchymal damage.^5,6 In cardiovascular disease, modulation of ferroptotic signaling confers protection against myocardial ischemia-reperfusion injury and aortic dissection.^7,8

To contextualize ferroptosis-inducing strategies within the broader oncology landscape, we first briefly compare the principal physical and immunological modalities currently in clinical or late-preclinical development. Beyond ferroptosis-specific approaches, the broader landscape of cancer treatment has been reshaped substantially by advances in physical and immunological modalities. Phototherapy, including photodynamic and photothermal variants, uses light-activated compounds to generate reactive oxygen species (ROS) or localized hyperthermia for tumor ablation. It delivers high spatiotemporal precision, yet tissue penetration is typically limited to less than one centimeter, and reliance on oxygen supply blunts its efficacy in hypoxic tumor cores. Sonodynamic therapy bypasses this depth limitation by using ultrasound to activate sensitizing agents, reaching up to approximately 10 cm into soft tissue with minimal off-target effects and no phototoxicity, though the lack of standardized clinical dosimetry remains a major barrier to translation. Immunotherapy, most notably immune checkpoint blockade, has transformed clinical outcomes by reinvigorating endogenous antitumor immunity, but responses are restricted to a subset of patients with pre-existing immune-infiltrated “hot” tumors, and both primary and acquired resistance are frequently observed.^9,10 Each of these modalities offers distinct advantages for specific clinical scenarios, yet all are constrained by tumor-microenvironmental factors such as hypoxia, limited tissue penetration, or insufficient immune infiltration—precisely the conditions under which ferroptosis inducers may provide complementary benefit. A key distinguishing feature of ferroptosis-inducing nanotherapeutics is their capacity to act as a synergistic partner for all these modalities: iron-driven lipid peroxidation triggers a distinct form of immunogenic cell death that can amplify responses to phototherapy, sonodynamic therapy, and immune checkpoint blockade alike. Representative recent strategies include photooxidation-initiated biomimetic nanoreactors that couple phototherapy with immune reinvigoration, sequential magneto-activation platforms that integrate mild thermal, oxidative, and immune effects, and ultrasound-based sonodynamic approaches that overcome the tissue-penetration limits of light.^11–13

Despite this broad therapeutic potential, no agent has received regulatory approval specifically for ferroptosis-based cancer treatment. Multiple interrelated roadblocks stand in the way of translating preclinical ferroptosis research into clinical practice. Candidate agents generally display subpar pharmacokinetic traits marked by poor water solubility, rapid metabolic clearance and negligible native tumor affinity, while single-pathway inhibition easily gives rise to therapeutic resistance. Researchers also face a scarcity of small-molecule modulators that balance potent target activity with mild systemic side effects, alongside a critical lack of non-invasive imaging tools capable of real-time tracking target engagement and therapeutic outcomes.^14,15 These four bottlenecks were identified through a systematic review of preclinical ferroptosis literature and registered clinical trials, focusing on recurring translational barriers common across small-molecule, nanomedicine, and biologic modalities, rather than platform-specific limitations. This framework maps directly onto the four bioengineering intervention strategies that structure the remainder of the review, in the order in which the strategies address them, allowing consistent evaluation of how each approach addresses a specific translational barrier. The list is not intended to be exhaustive; rather, it reflects the most widely acknowledged obstacles in peer-reviewed commentaries and regulatory analyses of ferroptosis-targeted therapeutics.

Engineered delivery systems offer meaningful improvements in the solubility, stability, and tumor-specific accumulation of ferroptosis-modulating agents.¹⁶ Multi-pathway combinatorial nanoplatforms counteract resistance by simultaneously blocking compensatory antioxidant axes, or by pairing ferroptosis induction with other cell death programs and standard-of-care therapies.¹⁷ Stimuli-responsive platforms add further spatiotemporal precision to ferroptosis regulation.¹⁸ Theranostic probes combine therapeutic functions with real-time imaging modalities to fill the long-standing gap in treatment monitoring. Representative examples include (4S)-4-(3-[¹⁸F]fluoropropyl)-L-glutamate PET, redox-responsive gadolinium MRI agents designed to detect lipid peroxidation, and near-infrared fluorescent probes.^19,20

Bioengineering, which integrates materials science, nanotechnology, gene editing, and stimuli-responsive design, provides a versatile set of tools to tackle these bottlenecks.²¹ Prior reviews have cataloged ferroptosis nanoplatforms by material class or molecular target. What is still missing, and what a translational team actually requires in practice, is a framework for deciding which engineering problem to address first. This review is organized around four complementary bioengineering strategies: targeted delivery to individual ferroptosis regulatory nodes, multi-pathway combinatorial synergy, smart stimuli-responsive platforms, and integrated theranostic systems. For each strategy, we examine both its underlying mechanistic rationale and its translational limitations, distinguishing challenges unique to the approach from broader obstacles shared across the nanomedicine field. We evaluate how each strategy addresses the four core bottlenecks, offer a critical reassessment of directions with uncertain clinical value, and outline actionable pathways to advance clinical translation.

These strategies correspond directly to the four bottlenecks identified above: pharmacokinetic limitations are addressed by Level 1 delivery systems; resistance by Level 2 combinations; selectivity by Level 3 responsive gating; and monitoring by Level 4 theranostic integration. Literature was identified through systematic searches of PubMed and Web of Science up to June 2026 using keyword combinations including ferroptosis, nanomedicine, cancer therapy, bioengineering, drug delivery, combination therapy, and stimuli-responsive systems. Only peer-reviewed original research and review articles published in English were included; conference abstracts, preprints, and non-English publications were excluded. Studies were screened by title and abstract, with full-text review for final inclusion, and reference lists of key reviews were cross-checked for additional primary studies. Regulatory guidance documents and clinical-trial registry records were cited directly as primary sources where no peer-reviewed equivalent exists. Representative platforms were selected based on the availability of peer-reviewed in vivo efficacy data, mechanistic distinctiveness, and proximity to clinical development. The clinical translation potential ratings in the tables are assigned by explicit, evidence-based criteria defined in the corresponding table footnotes.

Core Nodes and Engineering Intervention Strategies for Ferroptosis Therapy

Core ferroptosis nodes are defined as molecules that directly govern iron-dependent lipid peroxidation or intracellular antioxidant defense.^14,22 This category includes two classes of targets: those with validated, engineered intervention tools; and those that occupy irreplaceable positions in the ferroptosis execution cascade despite limited or preliminary engineering exploration.¹⁶ This definition distinguishes direct functional effectors from indirect upstream regulators and establishes a consistent framework for engineering-oriented therapeutic design.^23,24

Core Functional Classification of Ferroptosis Regulatory Nodes

Ferroptosis is regulated by three functionally linked classes of biological nodes, centered on iron metabolism, lipid peroxidation and antioxidant defense respectively.

Iron metabolism forms the foundational regulatory layer, revolving around the intracellular labile iron pool that supplies ferrous iron to drive Fenton reactions. Cellular iron uptake is mainly mediated by transferrin receptor 1 (TfR1), which delivers iron into cells via endocytosis followed by DMT1-mediated Fe²⁺ export from endosomes to the cytosol, while excess free iron is sequestered in ferritin complexes.^25,26 Nuclear receptor coactivator 4 (NCOA4) triggers ferritinophagy to release stored iron, expanding the labile iron pool and further amplifying Fenton-driven lipid peroxidation.^27,28

At the execution level, lipid peroxidation drives the final lethal damage to cell membranes. Polyunsaturated fatty acids (PUFAs) on cell membranes act as the primary substrates for peroxidation. Acyl-CoA synthetase long-chain family member 4 (ACSL4) catalyzes PUFA activation and integration into membrane phospholipids, and lipoxygenases (LOXs) then initiate peroxide formation on these membrane-incorporated PUFAs.^29–31

Against this oxidative damage, cells deploy a multi-layer antioxidant defense network to counteract lipid peroxidation. The SLC7A11–GSH–GPX4 axis serves as the canonical defensive pathway: solute carrier family 7 member 11 (SLC7A11) imports cystine to support glutathione (GSH) synthesis, and glutathione peroxidase 4 (GPX4) directly scavenges lipid peroxides to block ferroptosis.³² Two glutathione-independent auxiliary axes compensate when GPX4 activity is impaired: ferroptosis suppressor protein 1 (FSP1) reduces coenzyme Q10 (CoQ₁₀) to its membrane-embedded antioxidant form (CoQ₁₀H₂) at the plasma membrane, while dihydroorotate dehydrogenase (DHODH) does the same within mitochondria.^33–35 Mitochondrial membranes are themselves susceptible to lipid peroxidation, against which the DHODH–CoQ₁₀H₂ axis provides local defense.³³ The transcription factor nuclear factor erythroid 2-related factor 2 (Nrf2) further broadens antioxidant capacity by upregulating a panel of ferroptosis-defense genes.³⁶ A third route, the GTP cyclohydrolase 1 (GCH1)–tetrahydrobiopterin (BH₄) axis, remains less explored as a therapeutic target.³⁷

GPX4 and SLC7A11 represent the most well-validated intervention targets, with a wide range of nanocarrier formulations already tested in preclinical and early clinical settings.^38,39 Modulating the labile iron pool is another widely explored strategy, though long-term metal accumulation in the reticuloendothelial system brings unique biosafety concerns.⁴⁰ FSP1 and DHODH have emerged as promising complementary targets for combination regimens, while their druggability and target specificity still require further validation.³⁴ This functional ranking of nodes helps frame the rationale behind different nanomedicine design choices. The six principal intervention nodes and their spatial relationships are summarized in Figure 1.

Figure 1.

Core ferroptosis machinery and points of bioengineering intervention. Extracellular iron enters via transferrin (TF)–transferrin receptor 1 (TfR1) or engineered iron nanoparticles (Fe-NP; Fe₃O₄ core, polymer/lipid shell, PEG Corona), which accumulate in tumors through enhanced permeability and retention (EPR)-mediated passive targeting or ligand-mediated uptake (eg, HA–CD44), feeding the labile iron pool (LIP) and Fenton chemistry, while three defense axes—SLC7A11 (System Xc⁻)–GSH–GPX4, FSP1/DHODH–CoQ₁₀/CoQ₁₀H₂, and GCH1–BH₄—restrain peroxidation of PUFA-containing phospholipids (PUFA-PLs); unrestrained plasma-membrane and mitochondrial lipid peroxidation converges on ferroptosis. Yellow lightning bolts mark engineered intervention points (inhibition or activation); red exclamation marks mark dose-limiting healthy-tissue risks (liver/spleen reticuloendothelial accumulation; renal tubular SLC7A11; broad multi-organ GPX4 expression). Imaging components: artemisinin–gadolinium (Art–Gd) MRI for labile iron (preclinical); ¹⁸F-FSPG PET for System Xc⁻ transport activity (investigational, not ferroptosis-specific). Solid arrows, biochemical flux; dashed arrows, cellular uptake or imaging readout; bar-headed lines, inhibition. The L1–L4 key (bottom) maps the depicted interventions onto the framework levels of Figure 2: L1, node-specific delivery (Fe-NP, SLC7A11, GPX4); L2, multi-pathway co-blockade (GPX4 + FSP1 + DHODH); L3, stimuli-responsive gating (External triggers, Acid-triggered release); L4, theranostic integration (Art–Gd MRI, ¹⁸F-FSPG PET). Created with Figdraw (Hangzhou Duotai Technology Co., Ltd., Hangzhou, Zhejiang, China; www.figdraw.com).

Abbreviations: ACSL4, acyl-CoA synthetase long-chain family member 4; BH₄, tetrahydrobiopterin; CD44, cluster of differentiation 44; CoQ₁₀, coenzyme Q₁₀; DHODH, dihydroorotate dehydrogenase; EPR, enhanced permeability and retention; FSP1, ferroptosis suppressor protein 1; GCH1, GTP cyclohydrolase 1; Glu, glutamate; GPX4, glutathione peroxidase 4; GSH, glutathione; GTP, guanosine triphosphate; HA, hyaluronic acid; MRI, magnetic resonance imaging; NCOA4, nuclear receptor coactivator 4; PEG, polyethylene glycol; PET, positron emission tomography; PUFA, polyunsaturated fatty acid; RES, reticuloendothelial system; ROS, reactive oxygen species; US, ultrasound.

Targeted Engineering Intervention and Practical Restrictions

Intervention Targeting SLC7A11/System Xc⁻

The cystine/glutamate antiporter System Xc⁻ is a heterodimer of the catalytic light chain SLC7A11 and the heavy-chain chaperone SLC3A2. It imports extracellular cystine in exchange for intracellular glutamate. Agents such as erastin, sulfasalazine (SAS), and HG106 induce ferroptosis by blocking SLC7A11‑mediated cystine uptake and disabling glutathione‑dependent antioxidant defense.^41–43 A defining constraint for clinical translation is the intrinsic safety liability imposed by SLC7A11 expression in healthy tissues. SLC7A11 is abundantly and constitutively expressed in renal proximal tubular epithelial cells, where it supports cysteine supply and redox balance during oxidative stress.³⁸ Systemic exposure therefore places renal tissue at inherent risk of on‑target ferroptotic stress, independently of tumor‑targeting design.

In vivo studies confirm that systemic erastin administration reduces SLC7A11 and GPX4 expression in healthy renal tissue, elevates lipid peroxidation, and alters glomerular and tubular morphology consistent with sustained oxidative stress.^44,45 This creates a fundamental design conflict: formulations that efficiently deliver drug to tumors will also expose the kidney to pharmacologically relevant inhibition. This mismatch cannot be overcome by passive accumulation through the enhanced permeability and retention (EPR) effect, because tissue exposure is governed by target expression rather than selective retention.

As a result, renal safety concerns persist even with nanocarrier‑based delivery, narrowing the therapeutic window and limiting dose escalation. For SLC7A11‑targeted ferroptosis therapy to advance, delivery systems must achieve tumor‑restricted pharmacology rather than merely enhanced accumulation.

Intervention Targeting GPX4

RSL3, ML210, and GPX4‑targeted small interfering RNA (siRNA) directly disable lipid peroxide clearance and trigger pronounced ferroptosis in many tumor types.^38,46,47 The central engineering challenge is control over drug release in vivo. Many GPX4 inhibitors are highly potent but poorly soluble and may induce systemic toxicity if released prematurely in the circulation. While stimuli-responsive micelles and liposomes have been developed to improve stability and tumor accumulation, such materials can undergo partial leakage during blood circulation, allowing free inhibitor to accumulate in metabolic organs including the liver.^48–50

This leakage reduces the proportion of drug reaching solid tumors and may contribute to systemic safety risks that constrain practical dosing. Many stimuli-responsive systems designed for robust activation in the tumor microenvironment face challenges in clinical translation related to batch-to-batch consistency and in vivo stability.⁴⁰ Further development therefore demands carriers with improved circulatory stability that release active GPX4 inhibitors under tight, tumor-specific conditions, minimizing premature exposure and maximizing intratumoral activity.⁵¹ For GPX4-targeted ferroptosis therapy, the translational barrier is therefore twofold: the target itself is not tumor-restricted, and the carrier cannot fully prevent premature leakage. These two constraints set biological and formulation ceilings that no single carrier design can fully circumvent. The safety implications of GPX4 inhibition are underscored by non-tumor toxicology studies. Ma et al demonstrated that piR-16404 drives ferroptotic liver injury via the CASTOR1/mTORC1/GPX4 axis in HepG2 cells and mice, establishing a direct mechanistic link between GPX4 dysregulation and hepatocyte ferroptosis.⁵² While this study addresses chemical-induced liver toxicity rather than anticancer therapy, it reinforces the need for stringent tumor-restricted GPX4 inhibition to avoid hepatic on-target toxicity.

Intervention Targeting Intracellular Labile Iron Pool

Iron‑based nanomaterials and artemisinin derivatives induce ferroptosis by expanding the labile iron pool and amplifying Fenton chemistry.^53–55 These agents benefit from established clinical translatability, as several iron formulations are already used clinically.⁴⁰ The principal limitation is spatial: intratumoral iron distribution and redox state vary considerably across regions. Hypoxic zones, poorly perfused areas, and necrotic cores respond inconsistently to systemic iron supplementation, producing uneven ferroptosis induction and incomplete tumor eradication.^32,56

Non‑targeted iron nanomaterials often accumulate in the liver and spleen, potentially reducing tumor‑directed efficiency.⁵⁷ Iron-driven peroxidation is strictly governed by local oxygen tension and redox balance; uniform dosing therefore cannot be expected to produce uniform effects across a heterogeneous tumor mass.

Engineering strategies must therefore move beyond passive accumulation to redox‑feedback or microenvironment‑responsive iron release, allowing iron delivery to better align with the metabolic state of individual tumor regions.⁵⁸ The core limitation of iron pool targeting is thus spatial: iron delivery is gated by local redox and oxygenation status, not by tumor presence alone. This constraint cannot be overcome by formulation alone.

Intervention Targeting FSP1 and DHODH Compensatory Pathways

Inhibitors such as iFSP1 and brequinar target the FSP1–CoQ₁₀ and DHODH pathways, which mediate resistance to GPX4-directed ferroptosis.^33–35 iFSP1 blocks the regeneration of CoQ₁₀H₂ at the plasma membrane, whereas brequinar inhibits the same reaction within the inner mitochondrial membrane via DHODH. Both axes are upregulated when GPX4 is impaired, providing parallel antioxidant defense that sustains tumor cell survival. Engineering strategies for these targets have so far concentrated on co-encapsulating small-molecule inhibitors, rather than modifying the inhibitors themselves, to overcome rapid clearance and off-target accumulation of free drug combinations.

Clinical advancement faces two linked challenges: suboptimal druggability of some tool compounds and the absence of tissue-selective delivery systems.⁵⁹ iFSP1 has limited metabolic stability in vivo despite potent cell-based activity. Brequinar, though clinically tested in AML, was terminated for insufficient monotherapy efficacy and myelosuppression. Co-delivery of both inhibitors alongside GPX4-directed payloads within a single nanocarrier is theoretically attractive, but three practical barriers persist. The optimal release kinetics of each agent differ substantially. The drug ratio within a single carrier cannot be maintained reliably across individual tumors. Whether all three inhibitors must simultaneously reach overlapping tumor regions in individual patients is an assumption that has not been verified in biopsy-derived samples.

Systemic co-administration of multiple antioxidant inhibitors raises safety concerns because normal tissues also rely on redundant antioxidant pathways to maintain homeostasis.³⁶ Direct in vivo evidence of coordinated toxicity remains limited, so the primary constraint is not documented toxicity, but the absence of mechanisms to confine dual or triple inhibition to the tumor. Free drug combinations act systemically and may not be safely dosed at levels required for sustained tumor suppression.^47,60

Progress requires tumor-targeted co-delivery platforms that concentrate combinations within the tumor microenvironment, allowing coordinated pathway blockade while minimizing systemic exposure.^4,61 The safety threshold for triple blockade is not simply additive. Even if each single agent is tolerated at its maximum dose, the combination may compress the therapeutic window to an unacceptable degree. This sets a quantitative performance bar for delivery systems. Technical feasibility of co-encapsulation alone does not constitute success. Carriers must achieve intratumoral inhibitor concentrations sufficient to meet synergistic thresholds, while reducing cumulative exposure in normal tissues by at least an order of magnitude. That requirement is substantially more stringent than for single-node delivery or two-drug combinations.

Intervention Targeting ACSL4 and Lipid Metabolism

PUFA-based nanoparticles and lipid-modulating formulations enhance ferroptosis by reshaping membrane lipid composition and elevating peroxidation potential.^62,63 ACSL4 remains a compelling target due to its central role in ferroptosis execution, yet no potent and selective ACSL4 inhibitor has been widely validated for in vivo use.^64,65 The available tool compounds remain preclinical, with limited selectivity and no reported tumor distribution data. Delivery strategies therefore bypass direct enzyme inhibition altogether and instead deliver lipid substrates or modulators that shift membrane composition toward a more peroxidation-prone state.

The main practical hurdles are relatively poor circulatory stability and non-specific uptake by lipid-rich healthy tissues.⁶⁶ PUFA-containing carriers are prone to oxidation in the bloodstream, which can reduce activity before reaching tumors. In addition, adipose tissue and other lipid-rich organs may sequester lipid-like nanoparticles, diluting tumor delivery and contributing to off-target exposure.⁶⁷ Biomimetic or membrane-coated carriers may improve stability and circulation time, but selective delivery to tumor tissues remains a key hurdle for consistent efficacy.⁶⁸

This is not a formulation problem that better coating can fix. The fundamental barrier is biological: the same lipid uptake and transport machinery that delivers PUFAs to tumors also delivers them to healthy tissues. Until a strategy capable of actively directing lipid cargoes away from adipose tissue and toward tumors is established, or until a clinically viable ACSL4 inhibitor enters clinical practice, PUFA-based ferroptosis sensitization will remain a pharmacologically attractive yet delivery-limited approach.

Limitations of Single-Node Targeting

Single-node ferroptosis induction often fails to trigger sustained tumor regression, even for agents that exhibit substantial efficacy in cell culture models.¹⁶ Three key interconnected factors undermine the clinical performance of single-node therapeutic strategies in patients, which are adaptive resistance, intratumoral heterogeneity and a narrow therapeutic window.

Tumor cells initiate compensatory mechanisms within hours to days after pathway blockade. They upregulate parallel antioxidant systems, predominantly the FSP1–CoQ₁₀ and DHODH pathways, to restore lipid peroxide clearance and evade the inhibitory effects of single-node targeting.^33,34 Such rapid feedback regulation shortens the effective duration of single-agent treatment and renders fixed-dose regimens difficult to achieve stable therapeutic efficacy. In addition to intrinsic tumor adaptation, insufficient drug delivery across tumor lesions further compromises treatment outcomes. Variable perfusion, oxygenation and nutrient levels in different regions of a single tumor lead to distinct ferroptosis sensitivity among tumor cell populations.³² Even when nanocarriers successfully accumulate in tumor tissues, dense stromal structures and elevated interstitial pressure confine their distribution to perivascular regions. Tumor cells in deep, poorly vascularized areas therefore remain largely insensitive to treatment.⁶⁹

The third limiting factor lies in the constrained therapeutic window. SLC7A11 and GPX4 are not exclusively expressed in tumors. They undertake critical redox regulatory functions in the kidney, liver and brain. Targeted inhibition of these proteins in normal organs establishes a maximum safe clinical dose. This permissible dose is frequently insufficient to reach the intratumoral concentration required for thorough tumor suppression.^38,45 The three aforementioned barriers function independently, and no monotherapy formulation can overcome all these obstacles simultaneously. This drives the field to advance multi-pathway combination therapies and stimuli-responsive delivery systems. Single-node inhibitors are not therapeutically ineffective. Instead, their anti-tumor activity is consistently counteracted by intrinsic tumor biological characteristics, physical delivery barriers and human pharmacological properties.

Upstream Regulators with Defined Engineering Value

Metabolic Checkpoints

The mevalonate pathway and glutaminase GLS1 act as upstream modulators of ferroptosis sensitivity.^54,55 The AMP-activated protein kinase (AMPK)/Nrf2 axis serves as another upstream regulator that enhances antioxidant defense against ferroptosis.⁷⁰ Clinically available agents including atorvastatin and CB-839 have been shown to prime tumor cells for ferroptosis, in part by disrupting membrane integrity or redox balance.^53,71 Beyond approved agents, natural products have also emerged as ferroptosis modulators in preclinical studies. The phosphoinositide 3-kinase (PI3K)/protein kinase B (AKT)/mechanistic target of rapamycin (mTOR) pathway has been identified as a metabolic checkpoint regulating ferroptosis sensitivity: chrysotoxine, a natural phenanthrene compound, induces ferroptosis in cervical cancer models through modulation of the p53/SLC7A11/GPX4 axis, demonstrating that pharmacologically targeting oncogenic signaling cascades can lower the ferroptosis threshold.⁷² A key engineering consideration is limiting systemic metabolic disruption: statins affect hepatic lipid metabolism, while CB-839 may alter skeletal muscle homeostasis. Thus, tumor-restricted delivery is required to uncouple therapeutic priming from systemic metabolic effects.^73–75 In addition to synthetic agents and pathway modulators, natural products have been explored as ferroptosis inducers. For instance, paeonol has been shown to inhibit glioma cell growth by inducing ferroptosis, as evidenced by increased lipid peroxidation, reduced GPX4 expression, and elevated intracellular iron levels—though this study examined the free compound rather than a nanocarrier formulation.⁷⁶ The natural product scaffold offers chemical diversity for ferroptosis modulator discovery, but its translation will require tumor-restricted delivery to overcome the lack of intrinsic tumor selectivity shared by most small-molecule agents.

Post-Translational Regulation

Ubiquitin-specific proteases including USP7 and USP8 modulate the stability of core ferroptosis regulators, indirectly shaping sensitivity.^77,78 These targets offer opportunities for PROTAC-mediated degradation or peptide-guided targeting.⁷⁹ However, few tools have been validated in vivo, and off-target effects on protein homeostasis remain poorly defined. Translation remains contingent on selective delivery and predictable pharmacology.^35,80

Post-transcriptional regulation of ferroptosis-defense genes has also been demonstrated in non-oncologic settings. Zhong et al showed that YTH N6-methyladenosine RNA binding protein 1 (YTHDF1), transcriptionally activated by transcription factor 4 (TCF4), accelerates GPX4 and SLC7A11 translation in osteoblasts, suppressing ferroptosis in titanium nanoparticle-induced osteolysis.⁸¹ Although this study addresses a non-cancer pathology, it provides mechanistic evidence that RNA-level regulation of the canonical GPX4 and SLC7A11 ferroptosis-defense pathways is therapeutically actionable—a principle that may extend to cancer contexts where YTHDF1 is also overexpressed.

Beyond reversible small-molecule inhibition, proteolysis-targeting chimeras (PROTACs) represent an emerging complementary strategy for direct degradation of ferroptosis-associated regulators. Proof-of-concept studies have reported PROTACs targeting FSP1 and SLC7A11; when delivered via stimuli-responsive nanocarriers, these constructs enable prolonged pathway suppression superior to conventional small-molecule inhibitors in preclinical models.^82,83 Still, they face identical translational bottlenecks, including insufficient tumor-specific delivery, non-specific off-target protein degradation, and unfavorable pharmacokinetic profiles of large chimeric molecules, with in vivo validation still lacking. Collectively, although PROTACs enrich the ferroptosis-targeting toolbox, they do not circumvent the fundamental demand for tumor-restricted delivery and activation.

Cancer Stem Cell Phenotypes

Cancer stem cells (CSCs) readily resist ferroptosis due to persistently activated antioxidant programs: they maintain high GPX4 levels and robust Nrf2 signaling, alongside limited labile iron storage, allowing them to survive oxidative stress and drive tumor recurrence after treatment.^16,84 Ligand-functional nanoparticles (hyaluronic acid-coated carriers, for instance) targeting CSC surface markers CD44 and CD133 can selectively deliver erastin or iron nanoagents to aldehyde dehydrogenase (ALDH)-positive stem subpopulations, partially eradicating resistant CSC pools.⁸⁵ The biological characteristics of CD44 further rationalize this targeting design. This surface receptor stabilizes SLC7A11 to sustain intracellular antioxidant signaling,⁸⁶ while CD44-dependent endocytosis exhibits preferential uptake of hyaluronic acid-conjugated iron species.⁸⁷ On this basis, hyaluronic acid-functionalized iron nanocarriers can exploit the endogenous uptake cascade unique to cancer stem cells to achieve selective intracellular delivery.

Hypoxia deepens this barrier through hypoxia-inducible factor 1-alpha (HIF-1α), which directly upregulates GPX4 and sustains System Xc⁻-mediated glutathione synthesis via hypoxia-induced circHIF1α.^88,89 Crucially, this mechanism is already a delivery target rather than immutable biology: biomembrane-coated nanoparticles targeting circHIF1α inactivate System Xc⁻ through SLC7A11/SLC3A2, inducing ferroptosis in ovarian cancer stem cells and suppressing metastasis and cisplatin resistance.⁸⁹ A second niche barrier is physical: dense extracellular matrix and high interstitial pressure block nanoparticle infiltration. Tumor-penetrating peptides and matrix-degrading enzymes have been explored to improve nanoparticle penetration into dense tumor stroma. The iRGD peptide, belonging to the cell-penetrating peptide class, targets αvβ3/αvβ5 integrins and neuropilin receptors commonly expressed in tumor cells, thereby enhancing tissue penetration and intratumoral accumulation.⁹⁰

Clinically, broad CD44 expression in healthy stem tissues creates notable off-target toxicity risks.⁹¹ Single CSC-targeted nanomedicines rarely produce durable remission. Rational combinations that remodel niche fibroblasts or reshape local immunity, paired with pH/ROS dual-responsive gated release systems, are required to confine ferroptosis activation exclusively within malignant lesions.^88,92 A representative integrated strategy involves biomineralized hyaluronic acid-modified amorphous calcium phosphate nanounits co-delivering the signal transducer and activator of transcription 3 (STAT3) inhibitor niclosamide and the ferroptosis inducer RSL3; this system achieves CD44-targeted CSC recognition, STAT3–programmed death-ligand 1 (PD-L1) axis inhibition to stimulate adaptive immunity, and interferon gamma (IFNγ)-mediated downregulation of SLC7A11/SLC3A2, collectively cooperating with RSL3-induced GPX4 deactivation to trigger pronounced ferroptosis in CSCs.⁹³

Three Overarching Translational Constraints

Beyond the field-wide challenges discussed in the preceding sections, ferroptosis-directed nanotherapeutics face three additional hurdles intrinsic to the iron-dependent execution mechanism itself, which distinguish them from other classes of nanomedicine.

Three fundamental translational barriers stand out. First, key effector proteins including SLC7A11 and GPX4 carry physiological roles beyond malignant tissue. They are consistently expressed in renal tubular cells and neurons, setting an intrinsic upper limit for safe dosing that no delivery formulation can fully overcome.⁹⁴ Second, passive tumor accumulation built around the EPR effect lacks reliability in human patients. Tumor uptake via this mechanism varies drastically between lesions and fails to reliably forecast therapeutic outcomes.^95,96 The third limitation is specific to iron-laden nanomedicines and stems from their distinct toxicokinetic profiles. Small-molecule ferroptosis modulators clear the body rapidly, but iron nanoparticles lodge in the liver and spleen for months.^95,97 At present, there remains no unified preclinical safety standard to assess long-term risks from sustained metal tissue deposition.

Many cancer patients present with baseline iron dysregulation, including anemia of chronic disease, iron overload syndromes, and therapy-induced iron deposition. Systemic administration of iron-based ferroptosis inducers may exacerbate these pre-existing conditions.⁵¹ To date, baseline serum ferritin, transferrin saturation, or liver iron content have not been systematically incorporated as enrollment criteria in published reports of ferroptosis-directed clinical trials. The absence of iron-specific stratification criteria distinguishes ferroptosis nanotherapeutics from other nanomedicines and represents a translational barrier that must be addressed before first-in-human trials can be rationally designed.

The Four-Level Framework: Definitions and Ordering Rationale

The strategies discussed in this review fall into four tiers. Level 1 comprises node-specific delivery systems: lipid, polymeric, or inorganic carriers that formulate ferroptosis modulators for improved solubility, stability, and tumor accumulation.⁹⁸ Level 2 platforms incorporate multiple payloads or activities to co-target parallel antioxidant axes or to combine ferroptosis with other cell death programs and conventional therapies.⁹⁹ Level 3 adds a gating function—pH-, ROS-, or hypoxia-responsive chemistries that confine drug release to the tumor microenvironment, or exogenous triggers such as light, ultrasound, or magnetic fields that enable on-demand activation.^18,100 Level 4 further appends an imaging moiety, allowing the same carrier to report target engagement and guide subsequent dosing.¹⁰¹ The four-level framework and its mapping to the four translational bottlenecks are illustrated in Figure 2.

Figure 2.

The four-level bioengineering framework mapped to the four translational bottlenecks, with the staged translation roadmap. Each bottleneck (top) is assigned to the level that primarily resolves it (middle): Level 1, node-specific delivery (single warhead against one core node; EPR-based passive or ligand-targeted carriers; lead candidate CNSI-Fe(II)); Level 2, multi-pathway combinations (antioxidant co-blockade, multi-death co-induction, ICD–immunotherapy synergy); Level 3, stimuli-responsive gating (endogenous triggers: pH, ROS, hypoxia, enzymes; exogenous: light, ultrasound, magnetic field); Level 4, theranostic integration (Fe₃O₄ MRI; ¹⁸F-FSPG PET). Levels are ordinal, not ranked by merit: each builds upon the lower ones, and increasing integration entails higher manufacturing complexity and regulatory risk. Bottom: translation roadmap—Stage 1, locoregional Level 1 platforms; Stage 2, prospectively validated companion imaging to unlock systemic Level 2 therapy; Stage 3, standardized good laboratory practice (GLP) toxicology gating systemic Level 3 platforms. Created with Figdraw (Hangzhou Duotai Technology Co., Ltd., Hangzhou, Zhejiang, China; www.figdraw.com).

Abbreviations: CNSI, carbon nanoparticle–iron(II) complex; EPR, enhanced permeability and retention; ICD, immunogenic cell death; MRI, magnetic resonance imaging; PET, positron emission tomography; PK, pharmacokinetics; ROS, reactive oxygen species; TME, tumor microenvironment; US, ultrasound.

The ordering of these tiers is not a value judgment. It reflects engineering composition: each higher level builds upon the lower ones. A stimuli-responsive liposome is still a liposome; a theranostic construct typically begins with a responsive carrier and adds a label; a two-drug combination presupposes a carrier capable of co-encapsulation. No strategy can bypass the formulation problems solved at Level 1. Two implications follow. First, the ordering is ordinal rather than evaluative. Integration increases manufacturing complexity and regulatory burden—carbon nanoparticle-iron(II) complex (CNSI-Fe(II)), the most clinically advanced ferroptosis nanomedicine to date, is a Level 1 platform.¹⁰² A higher number is not a synonym for “better”. Second, because risk rises with integration, clinical translation should begin with the simplest tier and advance only as human data de-risk the next. The mapping between tiers and bottlenecks is deliberate. Level 1 delivery systems answer the pharmacokinetic bottleneck, improving the exposure of agents that cannot reach tumors on their own. Level 2 combinations answer the resistance bottleneck, disabling the compensatory pathways that defeat single-node blockade. Level 3 stimuli-responsive platforms answer the selectivity bottleneck: because no ferroptosis modulator is intrinsically tumor-specific, gating release to the tumor milieu supplies spatially the selectivity that pharmacology alone has not achieved. Level 4 theranostic systems answer the monitoring bottleneck, reporting target engagement in the absence of validated biomarkers. Each tier touches the remaining bottlenecks secondarily, but the primary correspondence is one-to-one. Applied retrospectively, the framework classifies the platforms discussed in this review without ambiguity. CNSI-Fe(II) remains Level 1 despite its clinical lead, because it carries a single payload with no gating or imaging function. Single materials with intrinsic secondary activity, such as Cu-TCPP(Fe) nanosheets, which suppress GPX4 and the FSP1–CoQ₁₀ axis through multi-enzyme-like catalysis, are likewise Level 1 in engineering terms, even though their biology mimics a combination.¹⁰³ Level 2 begins where separate payloads are co-delivered, as in hydrogels co-loading CuFe₃O₄ nanoclusters with artesunate to engage ferroptosis and cuproptosis simultaneously.¹⁰⁴ Gated versions of the same payloads, including AND-gate dual-responsive constructs, are Level 3, and imaging-bearing constructs such as Fe₃O₄-based MRI theranostics are Level 4. The deciding variable is the highest level of engineering integration, not the number of biological effects.

Safety Profile Across Intervention Tiers

No single evaluation metric can fully define the safety landscape of ferroptosis-inducing nanotherapeutics, as toxic risks arise from multiple interwoven mechanisms that require parallel assessment. At the fundamental biological level, inherent safety liabilities stem from the intrinsic expression patterns of core therapeutic targets. GPX4 and SLC7A11 are not uniquely expressed in tumor tissue; they carry essential physiological functions within renal tubular epithelial cells, neurons and other healthy organs. Even optimized delivery vehicles cannot completely eliminate the risk of on-target off-tumor damage triggered by pathway suppression in normal tissue.^38,40 This creates an inherent biological upper limit for therapeutic windows, independent of any carrier formulation design. For SLC7A11 inhibitory regimens, kidney injury stands as the primary factor restricting maximum tolerable dosage, while GPX4 suppression brings broad toxic risks across multiple vital organ systems.^44,45

Beyond target-originated toxicity, nanocarrier-derived adverse effects create an additional layer of clinical risk. Most non-degradable metallic nanomaterials become trapped within the reticuloendothelial system (liver, spleen and bone marrow) for weeks or longer after systemic administration. Long-term metal buildup generates persistent oxidative stress, yet comprehensive toxicology data across large animal species remains insufficient to quantify such chronic hazards.^32,40 For iron-based ferroptosis nanomedicines, the catalytic reactivity that drives tumor lipid peroxidation also poses a hidden threat: retained iron in healthy organs may sustain chronic oxidative lesions, which standard short-term toxicity assays often fail to capture.⁴⁰ Copper-containing nanoplatforms face analogous concerns with even sparser long-term biodistribution data available. CuO nanoparticles trigger dose-dependent ROS overproduction and tissue oxidative stress, yet few systematic safety evaluations have been conducted for copper ferroptosis carriers to date.^74,104,105

Material biodegradability directly dictates how long nanomaterials linger inside tissues, and whether induced organ damage can recede over time. Ferritin nanocages and biodegradable polymeric carriers clear from tissues far faster than non-degradable inorganic nanoparticles, bringing lower long-term toxic potential, though their capacity for sustained drug release is comparatively limited. Metal-organic frameworks (MOFs) sit in an intermediate category: certain subtypes display favorable degradability and mild toxicity, yet toxicological comparisons confirm clear disparities based on central metal ion, with copper and manganese-based MOFs ranking highest in organ toxicity risk.¹⁰⁶ Such wide variability makes it impossible to draw universal safety conclusions for all MOF ferroptosis platforms. Biomimetic membrane-coated nanocarriers carry a distinct set of translational barriers. While they effectively evade immune clearance in preclinical rodent models, membrane protein batches show inconsistent composition, and inherent immunogenic risks in human subjects remain poorly characterized—a drawback absent from fully synthetic delivery systems.¹⁰⁷

Immune-related adverse effects remain an understudied dimension for ferroptosis nanomedicines, yet they cannot be overlooked in clinical design. Ferritin nanocages rely on endogenous protein scaffolds for tumor homing, but pre-existing anti-ferritin antibodies accelerate nanoparticle clearance and may trigger hypersensitive responses upon repeated administration.^108,109 Cell membrane-cloaked carriers introduce patient-specific antigen variability that complicates uniform safety prediction. Meanwhile, ferroptosis itself generates potent inflammatory damage-associated molecular patterns (DAMPs) to boost anti-tumor immunity; unchecked widespread ferroptosis across healthy tissue could trigger systemic inflammatory complications that offset therapeutic gains.^110,111

Safe dosing regimens cannot be separated from material clearance characteristics. Systemically injected metal nanoparticles exhibit prolonged circulation and tissue retention, so total cumulative organ burden across multiple treatment cycles matters far more than single injection dosage alone. The doses required to overwhelm liver sequestration in mice do not translate to human patients.¹¹² For any non-degradable metal nanoplatform, 90-day repeated-dose Good Laboratory Practice (GLP) toxicology assessments using two animal species should become the default regulatory expectation to define safe cumulative exposure limits. At present, no ferroptosis-targeted nanocarrier has completed this regulatory-grade toxicology dataset.⁴⁰

Another unresolved safety question centers on the reversibility of ferroptotic injury in healthy parenchymal tissue. Unlike apoptosis, a silent, programmed cell death pathway, ferroptosis is driven by self-amplifying lipid peroxidation cascades that may continue even after therapeutic agents are cleared from tissue. Little systematic research tracks how renal tubule and liver tissue recover after GPX4 or SLC7A11 inhibition, leaving clinicians without clear guidance on safe treatment intervals.^40,45 Without distinguishing transient, recoverable injury from permanent cumulative organ damage, dose scheduling can only rely on empirical preclinical observation rather than defined mechanistic evidence.

Existing safety evidence varies drastically across different material classes and therapeutic targets. Table 1 separates on-target organ toxicity from carrier-derived off-target damage, and marks the maturity of supporting clinical evidence for each core regulatory node, including three tiers: rodent-only preclinical results, early human exploratory trials (taking CNSI-Fe(II) as an example, with Phase I data from locoregional administration in Chinese cohorts and a Phase Ib/IIa trial registered), and known toxic profiles from unrelated clinical agents (such as brequinar’s myelosuppressive effects, which have not been validated for ferroptosis treatment). This sorting framework does not cover every potential toxic risk, but delivers a unified evaluation standard to contrast evidence sufficiency across different ferroptosis intervention strategies, clearly distinguishing well-supported safety conclusions from untested hazards that demand further preclinical and clinical validation before regulatory submission.

Table 1.

Core Ferroptosis Regulatory Nodes, Engineering Strategies, and Safety Risks

Target	Function	Modulators and Inhibitors	Delivery Platform	Key Safety Risk	Highest Evidence Level (Model Tier)
GPX4	Reduces lipid hydroperoxides	RSL3; GPX4-targeted siRNA47	Acid-activated nanoparticles (BNP@R);59 iron/siRNA co-loaded ferritin nanocages (HFn@Fe/siGPX4)47	Broad healthy-tissue expression; renal, hepatic and neuronal injury upon systemic suppression44,45	Immunocompetent syngeneic murine models (BNP@R); murine xenograft (HFn@Fe/siGPX4)
SLC7A11	Cystine import for GSH synthesis	Erastin, sulfasalazine, HG10641–43	Platelet-membrane-coated sulfasalazine nanoparticles (Fe3O4-SAS@PLT)113	Renal proximal tubular injury upon systemic SLC7A11 blockade44,45	Metastatic 4T1 model, immunocompetent mice
Labile iron pool (Fe2⁺)	Drives Fenton chemistry and hydroxyl radical generation	CNSI-Fe(II);57 artemisinin derivatives114	Intratumoral CNSI-Fe(II);57 acid-degradable TA–Fe/ART@ZIF114	Hepatic and splenic iron retention40	Published Phase I clinical data (locoregional intratumoral) (CNSI-Fe(II))102,115
FSP1	CoQ10-dependent lipid peroxide defense	iFSP134,35	Liver-targeted MOF co-loaded with RSL3 (RF@LA-Fe-MOF)61	Toxicity not yet characterized	Preclinical murine models
DHODH	Mitochondrial CoQ10H2 regeneration	Brequinar33	No validated targeted carrier (free drug only; a brequinar-loaded AND-gate system remains preclinical)116	Myelosuppression (single-agent risk)	Preclinical murine models
ACSL4	PUFA activation for ferroptosis sensitization	No clinically compatible in vivo inhibitor64,65	PUFA-enriched biomimetic nanovesicles (FiFe@RBM)63	Not applicable (no direct inhibitor)	Preclinical murine models

Note: This table summarizes core ferroptosis-regulatory nodes with proven bioengineering feasibility. Auxiliary defense pathways (eg, the GCH1–BH₄ axis) are excluded because no engineered delivery system currently addresses them. Safety risks are preclinically derived; clinical safety data remain unestablished except where indicated. Evidence levels denote the highest model tier reported in the cited primary studies. Citations are placed within the cell whose statement they support.

Abbreviations: ACSL4, acyl-CoA synthetase long-chain family member 4; ART, artemisinin; BH₄, tetrahydrobiopterin; CNSI-Fe(II), carbon nanoparticle–iron(II) complex; CoQ₁₀, coenzyme Q₁₀; DHODH, dihydroorotate dehydrogenase; FSP1, ferroptosis suppressor protein 1; GCH1, GTP cyclohydrolase 1; GPX4, glutathione peroxidase 4; GSH, glutathione; HFn, heavy-chain ferritin; MOF, metal-organic framework; PLT, platelet membrane; PUFA, polyunsaturated fatty acid; SAS, sulfasalazine; siRNA, small interfering RNA; SLC7A11, solute carrier family 7 member 11; TA, tannic acid; ZIF, zeolitic imidazolate framework.

Node-Specific Delivery Systems

Node-specific delivery is the field’s direct answer to the pharmacokinetic bottleneck—and its most basic and widely explored strategy: improving the solubility, stability, pharmacokinetic behavior, and tumor-targeting accumulation of ferroptosis modulators. Although single-node interventions cannot fully overcome therapeutic resistance or intratumoral heterogeneity, they establish the fundamental toolbox and design principles for the more sophisticated multi-pathway and smart responsive systems discussed in subsequent chapters.

A diverse nanocarrier toolbox has been developed for ferroptosis modulation, ranging from lipid and polymeric systems to inorganic frameworks, metal-based particles, and biomimetic or protein-based carriers.^16,59,68

Iron-Targeted Delivery Systems

Iron-based ferroptosis nanomedicines represent the only category with available clinical human research data. CNSI-Fe(II) serves as the leading candidate in this field. It is a carbon-iron composite capable of producing MRI signals and inducing intratumoral lipid peroxidation simultaneously.⁵⁷ A first-in-human clinical trial enrolled 19 patients with advanced solid tumors.¹¹⁵ Imaging results confirmed tumor central necrosis after treatment, and four heavily pretreated patients who had failed multiple prior therapies achieved long-term survival benefits.^102,117 This therapeutic approach faces a major bottleneck in delivery mode. CNSI-Fe(II) requires intratumoral injection, which limits its application to only superficial and accessible tumor lesions.¹¹⁵ The drawback does not stem from iron-based chemical properties. It is a common defect of systemic iron delivery. Free iron can be rapidly eliminated by the reticuloendothelial system, and intravenously administered iron nanoparticles tend to accumulate in the liver and spleen instead of targeting tumor tissues.⁴⁰

Researchers have engineered iron cores with auxiliary functions to enhance their catalytic performance and tumor-targeting specificity. Manganese-doped Fe₃O₄ exhibits pH-dependent dual enzyme-mimicking activity. It functions as a peroxidase and catalase to produce reactive oxygen species (ROS) in the acidic tumor microenvironment, while remaining biologically inert under normal physiological pH conditions. Another engineered formulation, FePt–glucose oxidase (GOx), depletes intratumoral glucose and generates hydrogen peroxide to sustain local Fenton reactions. Both design strategies effectively boost ROS production per iron atom. However, neither approach resolves the core obstacle of in vivo iron distribution. Only a small and variable proportion of delivered iron can accumulate in tumor tissues, while the majority accumulates in the liver.^66,67

Artemisinin and its derivative dihydroartemisinin are not iron carriers but iron-activated prodrugs. They require Fe²⁺ to generate cytotoxic radicals, making them dependent on the labile iron pool rather than expanding it. Nanoformulations such as TA–Fe/ART@ZIF release Fe²⁺ and artemisinin together in acidic endosomes, theoretically increasing radical production at the target site.¹¹⁴ The limitation is spatial heterogeneity: solid tumors contain hypoxic, poorly perfused regions with insufficient labile iron, so artemisinin activation remains uneven even when nanocarriers reach the lesion.⁵⁶ This mirrors the core barrier of direct iron delivery, where iron bioavailability, not carrier engineering, sets the rate-limiting step.

A spatial mismatch runs through all iron-based systems: iron delivery must align with local redox conditions, not merely with tumor presence. Redox-feedback or microenvironment-responsive release could in theory address this gap, but such approaches remain preclinical and untested in human tumors. For now, the most clinically advanced iron-based ferroptosis platform also carries the narrowest clinical indication, confined to locoregional rather than systemic use. This trade-off between translational maturity and clinical scope appears consistently across the ferroptosis nanomedicine field.

Antioxidant-Defense-Targeted Systems

Inhibiting System Xc⁻

Small-molecule inhibitors targeting System Xc⁻, including erastin and sulfasalazine, have been extensively validated to drive ferroptotic cell death. Their translational progress, however, is held back by weak in vivo bioavailability and poor tumor versus normal tissue selectivity.^4,118 One representative design leverages Fe₃O₄ nanoparticles wrapped with platelet membrane coatings, named Fe₃O₄-SAS@PLT; the biomimetic shell helps the carrier evade immune clearance and preferentially accumulate at metastatic tumor foci.¹¹³ Even with homotypic targeting capacity, this design cannot overcome the renal safety ceiling imposed by proximal tubular SLC7A11 expression; tumor-restricted activation remains the only path to a meaningful therapeutic window for this target class.¹¹⁹ We regard such repurposed small molecules as the most undervalued Level 1 cargo: their human safety records are already written, and only the delivery problem remains.

Depleting Glutathione

Researchers have developed two separate routes to cut intracellular glutathione pools down. The first suppresses GSH synthesis using buthionine sulfoximine (BSO), while the second relies on nanozyme catalysts that mimic glutathione oxidase to consume existing GSH.⁴² Biomimetic gold nanocages (m@Au-D/B) serve as a typical example, co-loading BSO and doxorubicin to simultaneously block glutathione production and raise intracellular oxidative load.¹²⁰ Another multi-enzyme platform, ZnFe₂O₄@Pt@PEG (ZFPG), consumes GSH under ultrasonic stimulation to achieve similar oxidative disruption.¹⁹ Both types of construction face identical translational hurdles: systemic GSH depletion disrupts redox balance across normal organs, and there remains no clear safety margin to separate tumor-specific oxidative damage from widespread off-tissue redox perturbation.¹²¹

Inhibiting GPX4

Direct GPX4 suppressors such as RSL3 produce potent ferroptosis effects in cultured cells, yet poor water solubility and systemic toxic liabilities hinder further in vivo translation.⁴⁷ Acid-responsive BNP@R is engineered to release loaded RSL3 only after internalization into acidic endosomes, avoiding premature drug leakage during circulation.⁵⁹ Ferritin nanocages (HFn@Fe/siGPX4) represent another viable design, co-encapsulating iron species and GPX4-targeted small interfering RNA to couple Fenton oxidative stress with genetic silencing of antioxidant machinery.⁴⁷ While both carriers boost tumor selectivity relative to free small-molecule agents, broad GPX4 expression across healthy tissues creates unavoidable risks of systemic cytoprotective pathway suppression, demanding strict tumor-localized activation for clinical safety.^38,44 Carriers can reduce drug exposure in healthy organs, but they cannot change where GPX4 is expressed; for this target class, tumor-restricted activation remains the ceiling set by biology, not by formulation.

PUFA Delivery for Lipid Peroxidation

Delivering exogenous polyunsaturated fatty acids (PUFAs) can raise membrane peroxidation susceptibility and sensitize tumor cells toward ferroptotic stimuli.¹²² Lipid nanoparticle formulation AA-LNP adjusts the intracellular ratio of polyunsaturated to monounsaturated fatty acids to amplify lipid breakdown triggered by oxidative stress.⁶² Biomimetic FiFe@RBM nanovesicles offer an alternative approach, reshaping cellular lipid metabolism while simultaneously boosting Fenton reaction efficiency.⁶³ Combining either carrier with stearoyl-CoA desaturase 1 (SCD1) inhibitors generates synergistic ferroptosis outcomes, yet both strategies struggle to precisely confine peroxidation damage to tumor tissue alone.

Beyond implementation challenges for specific formulations, PUFA-based sensitization faces a structural barrier that delivery engineering cannot circumvent: lipid peroxidation, once initiated, spreads across all lipid-bearing membranes regardless of malignant versus normal cell status. Carriers can reduce premature oxidation and modestly improve tumor-to-normal ratios, but they cannot render the peroxidation event itself tumor-specific.

From Single-Node Delivery to Combinatorial and Responsive Systems

Among all single-target systems, CNSI-Fe(II) is the only one that has reached human trials.^102,117 Systemic iron formulations face a delivery ceiling: only a median of 0.7% of injected nanoparticles reach solid tumors,⁶⁹ and even Kupffer-cell-saturating doses that yield up to 12% ID/g in mice¹¹² are unlikely to translate to humans. Still, interventions focused on only one regulatory node carry built-in functional limits. Tumor cells readily activate compensatory antioxidant signaling to counteract single-pathway suppression.^34,35 Heterogeneous perfusion and dense stroma prevent uniform nanoparticle distribution across a lesion, so ferroptosis induction remains patchy.^4,123 These are not failures of individual carriers; they are intrinsic to single-node logic. The field therefore builds on these delivery platforms to develop two complementary directions: multi-pathway carriers that block compensatory resistance axes together, and stimuli-responsive vehicles that restrict ferroptosis to the tumor microenvironment.

Comparative Performance Across Material Classes

Having detailed a spectrum of iron-targeted nanocarriers in the preceding subsections, it is meaningful to draw cross-category comparisons between iron-based nanocatalysts, copper-engineered therapeutic agents and conventional small-molecule ferroptosis triggers. Distinctions in their working mechanisms, in vivo pharmacokinetic profiles and overall translational readiness are far more than descriptive academic differences; they directly shape dosing regimens, long-term safety planning and the suitability of each material class for distinct tumor types.

Iron-containing nanomaterials stand furthest along the path toward clinical translation. Representative formulations including CNSI-Fe(II) and ferritin nanocatalysts leverage intrinsic iron-mediated Fenton chemistry to drive lipid peroxidation, while readily integrating extra functions such as MRI imaging or co-delivered gene silencing payloads.^47,124 A major strength of this class lies in established clinical precedent, with multiple iron supplements already approved for systemic use and CNSI-Fe(II) having completed a Phase I trial and being evaluated in a Phase Ib/IIa trial (NCT06048367 and NCT07433283).^102,115,117 Despite this, systemic iron administration faces persistent limitations: injected nanoparticles are rapidly captured by the reticuloendothelial system, leading to uneven drug distribution across tumor lesions. Slow iron clearance from hepatic and splenic tissue also creates lingering chronic toxic risks, and therapeutic efficacy fluctuates widely depending on local intratumoral redox conditions.⁴⁰

Copper-based nanoplatforms offer a unique mechanistic alternative to iron-centered designs. Under suitable biochemical environments, copper exhibits stronger catalytic activity in Fenton-like reactions, and its intracellular accumulation simultaneously initiates cuproptosis alongside ferroptosis to deliver dual cell death effects that iron-only systems cannot replicate.^104,125 CuFe₃O₄ injectable hydrogels serve as a typical localized formulation built around this synergistic logic.¹⁰⁴ Despite such theoretical strengths, copper nanomedicines lag far behind iron counterparts in translational progress. Mammalian organisms maintain tight homeostatic control over copper levels; unregulated systemic copper exposure readily triggers liver and neuronal damage with a narrower therapeutic window. Most existing copper nanosystems also lack complete long-term biodegradation and biodistribution data, slowing regulatory advancement.⁷⁴

Small-molecule ferroptosis activators such as erastin, sulfasalazine and RSL3 feature well-characterized target interactions and clear pharmacological readouts.^126–128 Sulfasalazine even carries existing clinical approval for unrelated inflammatory indications, opening avenues for drug repurposing work. Their core drawback comes from a lack of intrinsic tumor tropism. After systemic delivery, these small molecules freely circulate and suppress SLC7A11 or GPX4 across all healthy tissues where these proteins are expressed, bringing irreversible renal and hepatic off-target injury.^38,45 Low aqueous solubility and rapid metabolic clearance further narrow their effective therapeutic range. Unlike tunable nanocarriers, small-molecule structures cannot easily be reworked for controlled release or multimodal imaging, limiting their compatibility with timed combinatorial treatment plans.

No single material class is universally superior across all tumor indications; agent selection must be guided by real-world clinical context. For superficial, surgically accessible lesions where local delivery limits systemic exposure, iron-based nanomaterials remain the most mature and clinically practical option. Copper-based platforms hold particular promise for apoptosis- and ferroptosis-resistant tumors requiring dual metal-dependent cell death signaling, but require rigorous long-term toxicology validation prior to clinical translation. Small-molecule ferroptosis inducers remain viable for systemic use only when paired with validated companion imaging biomarkers to stratify responsive patient subgroups. Formal head-to-head in vivo comparisons across these three material classes have not been reported to date, so the relative prioritization outlined above rests on cross-study inference rather than direct experimental evidence, with each class occupying a distinct therapeutic niche shaped by tumor anatomy, intrinsic resistance profiles, and available clinical delivery infrastructure. All platforms discussed in this section fall under Level 1 of our four-tier engineering framework. This classification implies no hierarchical ranking: CNSI-Fe(II), the only ferroptosis nanomedicine with clinical data in humans, is itself a Level 1 construct, making locoregionally administered Level 1 platforms the most credible, fastest path to establishing human proof of mechanism.

Multi-Pathway Combination and Synergy

To overcome the resistance bottleneck created by single-pathway targeting, next-generation bioengineering strategies co-target multiple ferroptotic regulatory axes or combine ferroptosis induction with other therapeutic modalities. This section covers three core directions: co-targeting of parallel antioxidant defense pathways (GPX4, FSP1, DHODH); combinatorial pairing of ferroptosis with other regulated cell death programs (cuproptosis, pyroptosis, apoptosis); and synergistic combinations with conventional and emerging cancer therapies. Representative multi-pathway platforms and their translational limitations are summarized in Table 2. By definition, all constructs in this chapter are Level 2 designs. Classification is determined by the highest level of engineering integration; co-delivery of independent payloads, rather than the number of regulated biological pathways, forms the core qualifying standard.

Table 2.

Representative Multi-Pathway Ferroptosis-Combined Therapeutic Platforms and Translational Limitations

Combined Strategy	Representative Platforms	Core Synergistic Mechanism	Key Translational Limitation	Highest Evidence Level (Model Tier)
Dual ferroptosis-defense blockade (GPX4 + FSP1)	RF@LA-Fe-MOF;61 Cu-TCPP(Fe) MOF nanosheets103	Simultaneous inhibition of two parallel lipid-peroxide defense pathways	High formulation complexity; GMP manufacturing difficulty; undefined hepatic/renal toxicity	Preclinical murine models
Ferroptosis + immunotherapy	PD-1-displaying membrane-coated RSL3 nanoparticles;129 fucose-targeted DAPC/CpG liposomes130	Ferroptotic ICD activates antitumor immunity and synergizes with immune checkpoint blockade	PK–PD mismatch between DAMP release and antibody kinetics; lack of predictive biomarkers; T-cell ferroptosis-mediated immune exhaustion	Immunocompetent syngeneic models (4T1 triple-negative breast cancer)
Ferroptosis + chemotherapy	PSMA-targeted arsenic nanosheets loaded with doxorubicin (PMANs);131 cisplatin–RSL3 liposomes132	Chemotherapy-induced ROS and DNA damage amplify ferroptosis and overcome drug resistance	Overlapping off-target toxicity; poor tumor selectivity; mismatched release kinetics	Murine subcutaneous xenograft
Ferroptosis + radiotherapy	Intratumoral CNSI-Fe(II);57,133 inhalable biomimetic nanoreactor (DHA-N@M)134	Ionizing-radiation-driven ROS and SLC7A11 downregulation amplify iron-dependent lipid peroxidation	Unoptimized radiation-nanoparticle timing; unclear normal-tissue radiosensitivity; undefined dose fractionation	Published Phase I clinical data (CNSI-Fe(II)); orthotopic lung tumor model in immunocompetent mice (DHA-N@M)
Ferroptosis + energy-based therapy (PDT/SDT)	Ce6–erastin supramolecular nanoassembly (PDT);135 BTO/MoS2@CA sonocatalyst (SDT)136	External energy generates ROS and heat to trigger lipid peroxidation	Limited tissue penetration of light; non-standardized ultrasound dosimetry; bundled multi-effects hinder mechanistic verification	Preclinical murine models
Ferroptosis + gas/starvation therapy	Gas-donor nanoplatforms;137,138 glucose-oxidase–iron MOF cascade nanoreactors139	Gas-mediated redox disturbance or glucose depletion amplifies ferroptosis	Mechanistically uncoupled effects; off-target metabolic disturbance; low in vivo specificity	Preclinical murine models

Note: Evidence level denotes the highest model tier reported for each platform in the cited primary studies, in ascending order from subcutaneous xenograft to orthotopic to immunocompetent to clinical evidence. This tier is distinct from the engineering-level classification (Levels 1–4; Figure 2), which describes functional integration rather than validation maturity. CNSI-Fe(II) clinical data are derived from locoregional intratumoral administration in Chinese cohorts; these findings do not validate systemically delivered ferroptosis nanomedicine and should not be extrapolated to intravenous platforms.

Abbreviations: Ce6, chlorin e6; CNSI-Fe(II), carbon nanoparticle–iron(II) complex; CpG, cytosine-phosphate-guanine oligodeoxynucleotide; DAMP, damage-associated molecular pattern; DAPC, 1,2-diarachidonoyl-sn-glycero-3-phosphocholine; DHA, docosahexaenoic acid; FSP1, ferroptosis suppressor protein 1; GMP, good manufacturing practice; GPX4, glutathione peroxidase 4; ICD, immunogenic cell death; MOF, metal-organic framework; PD-1, programmed cell death protein 1; PDT, photodynamic therapy; PK–PD, pharmacokinetic–pharmacodynamic; PSMA, prostate-specific membrane antigen; SDT, sonodynamic therapy.

Co-Targeting Parallel Antioxidant Defense Axes (GPX4, FSP1, DHODH)

Compensatory activation of parallel antioxidant systems remains a central obstacle in ferroptosis therapy, driving the evolution of bioengineering strategies from dual-axis blockade to triple-axis blockade—co-targeting GPX4–GSH, FSP1–CoQ₁₀, and DHODH–CoQ₁₀H₂ within a single nanoplatform.^33,34 Of note, ACSL4 still lacks a clinically compatible in vivo small-molecule inhibitor, and the GCH1–BH₄ axis has not yet been addressed by engineered delivery systems.^34,37,64

Dual Inhibition of GPX4 and FSP1

GPX4 and FSP1 represent the two dominant and mutually compensatory antioxidant axes mediating ferroptosis resistance. Inhibition of one often induces upregulation of the other, making co-targeting the most straightforward and widely validated combinatorial strategy.

The liver-targeted MOF (metal-organic framework) system RF@LA-Fe-MOF co-encapsulates RSL3 (GPX4 inhibitor) and iFSP1 (FSP1 inhibitor), with surface lactobionic acid enabling asialoglycoprotein receptor-mediated hepatocyte targeting. Intracellular reduction of Fe (III) to Fe (II) drives Fenton chemistry and promotes controlled drug release. In liver cancer models, this dual-axis strategy suppresses tumor growth and lung metastasis, with combination index values below 0.7 confirming consistent synergy.⁶¹

Cu-TCPP(Fe) MOF nanosheets, which yield approximately 84% tumor growth inhibition in a murine breast cancer model, exert multi-enzyme-like activities to deplete GSH and NAD(P)H, thereby suppressing both GPX4 and the FSP1–CoQ₁₀ axis.¹⁰³ Together, these designs illustrate the potential of coordinated anti-ferroptosis pathway targeting, even as heightened complexity presents manufacturing and regulatory challenges. Beyond MOF-based platforms, inorganic nanoparticles offer an alternative strategy for GPX4/FSP1 co-targeting. Ma et al discovered that hollow mesoporous CuS nanoparticles (CuS NPs) possess an intrinsic capacity to inhibit GPX4; when loaded with iFSP1, these nanoparticles block both GPX4 and FSP1 antioxidant systems and cooperate with near-infrared irradiation to reinforce immunogenic cell death.¹⁴⁰ Unlike MOFs that rely on acidic degradation for drug release, CuS NPs combine intrinsic enzymatic activity with photothermal responsiveness, yet their non-biodegradable nature and potential long-term copper accumulation raise safety concerns that require careful evaluation.

Triple Axis Blockade (GPX4 + FSP1 + DHODH)

Triple-axis blockade extends dual-axis logic by adding DHODH inhibition, disabling the third parallel antioxidant system that remains intact even when GPX4 and FSP1 are suppressed, to overcome compensatory defense and drive sustained ferroptosis.

A representative example is the hollow iron-doped silica nanozyme (FeSHS), loaded with the DHODH inhibitor brequinar. This nanozyme exerts three synergistic enzyme-like activities, namely peroxidase, glutathione oxidase, and NAD(P)H oxidase, which collectively deplete GSH, impair GPX4 function, disrupt the FSP1–ubiquinone axis, and directly inhibit DHODH. In murine 4T1 breast cancer models, this multimodal platform achieves potent tumor growth inhibition.¹⁴¹

However, the individual contributions of each mechanism cannot be readily deconvolved given the parallel action of multiple enzymatic activities and pharmacological payloads. The inherent structural and functional complexity of this platform also creates major barriers to batch-to-batch reproducibility, quality control, and regulatory approval, positioning it as a sophisticated academic proof of concept rather than a clinically actionable candidate.

Repurposed marketed drugs offer an alternative strategy to sensitize tumors to ferroptosis by disrupting upstream metabolic and epigenetic pathways. Statins stand as a representative class: they block the mevalonate cascade to cut cellular CoQ₁₀ stores, dampening FSP1-mediated antioxidant defense and hindering the synthesis of selenoproteins such as GPX4.⁵⁵ This mechanism has been validated in nanoformulation studies. One design, simvastatin encapsulated within zwitterionic PCBMA-coated iron oxide nanoparticles, suppresses 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMGCR) to downregulate GPX4 while generating intracellular Fenton reactions, yielding marked anti-tumor effects in triple-negative breast cancer (TNBC) xenografts.¹⁴² Similarly, silk fibroin nanocarriers loaded with rosuvastatin were shown to reverse FSP1-driven ferroptosis resistance in the same tumor model.¹⁴³ Glutaminase inhibitor CB-839 limits glutamine supply for glutathione biosynthesis, indirectly reducing the glutathione pool available to GPX4 and boosting ferroptotic susceptibility.^53,144 Epigenetic agent tazemetostat modulates SLC7A11 and GPX4 expression via enhancer of zeste homolog 2 (EZH2) suppression, yet concrete data linking this activity to bona fide ferroptosis induction remains limited.^145,146 Notably, no nanomedicine construct has so far combined either compound with iron-based carriers, leaving room for further preclinical validation.

Repurposed therapies carry clear translational merits, as their pharmacokinetic behavior and safety risks in human patients are already well characterized to speed clinical progression. Their major downside stems from widespread target expression across healthy organs, meaning localized tumor delivery is mandatory to avoid off-tissue redox disruption. Under the four-tiered engineering classification laid out in this review, repurposed small molecules function best as auxiliary therapeutic cargo, for Level 1 and Level 2 nanosystems, rather than as standalone platform designs.

Challenges Specific to Intra-Network Co-Targeting

Despite clear synergy in vitro and in vivo, co-targeting multiple ferroptotic regulatory nodes introduces considerable translational complexities that must be addressed before clinical implementation.^14,32

A primary challenge is pharmacokinetic mismatch. Individual inhibitors often differ considerably in optimal release kinetics, chemical stability, and intratumoral penetration behavior. Encapsulation within a single nanocarrier does not ensure coordinated delivery or consistent molar ratios at the target site, highlighting the need for sequentially degradable structures or stimulus-activated cascade release systems.^20,22

Manufacturing scalability represents a second major hurdle. Multitargeted platforms with overlapping enzyme-like activities and multiple drug payloads are inherently more complex than single-agent formulations. Such systems frequently suffer from limited batch-to-batch reproducibility, poor storage stability, and challenging regulatory characterization, all of which impede compliance with Good Manufacturing Practice standards.³²

On-target off-tumor toxicity is amplified in multi-axis blockade: where single-node inhibition narrows the therapeutic window, simultaneous suppression of GPX4, FSP1, and DHODH in normal tissues (liver, kidney, proximal tubules) can effectively abolish it.^38,147 This places a premium on stringent tumor-restricted activation mechanisms, such as release triggered by acidity, ROS, hypoxia, or proteases, to confine combinatorial inhibition to malignant lesions.⁵⁹

A further unresolved need is effective targeting of the GCH1–BH₄ axis, which constitutes an additional compensatory antioxidant pathway. Pharmacological inhibition of this axis remains difficult to achieve in vivo, and no targeted nanoformulation has yet been reported. This gap represents a key opportunity for next-generation bioengineering development.^34,37 The GCH1–BH₄ pathway stands out as the most promising underexplored target within ferroptosis antioxidant networks. At the same time, triple inhibition targeting GPX4, FSP1 and DHODH comes with clear limitations. Adding a third inhibitor does not guarantee broader therapeutic efficacy, and researchers still lack solid evidence proving this triple combination can expand rather than shrink the therapeutic window.

Ferroptosis Combined with Other Regulated Cell Death Pathways

Combining ferroptosis with other regulated cell death pathways has become an effective strategy to overcome therapy resistance. These combinations follow three unifying principles: metal ion overload links ferroptosis and cuproptosis; ROS intensity determines whether cells undergo apoptosis, pyroptosis, or ferroptosis; and multi-mode combinations disable antioxidant defenses, damage membrane structure, and disrupt mitochondrial function simultaneously.

Ferroptosis + Cuproptosis

Ferroptosis and cuproptosis are both metal-dependent, converging on oxidative and metal overload to synergistically kill tumor cells.^104,125 A typical in situ hydrogel co-delivers CuFe₃O₄ nanoclusters and artesunate, elevating ROS and depleting GSH to trigger both ferroptosis and cuproptosis, effectively suppressing tumor growth in orthotopic osteosarcoma models.¹⁰⁴

Dual metal-based systems rely heavily on overlapping oxidative stress, making it difficult to quantify individual pathway contributions. Local hydrogel delivery reduces systemic metal exposure and EPR dependence, but coordinating two metal-dependent death programs increases mechanistic complexity and complicates safety validation.

Mechanistically, iron and copper jointly drive mitochondrial ROS generation, forming a self-amplifying oxidative cascade that sensitizes cells to both ferroptosis and cuproptosis.¹⁴⁸ This shared metabolic vulnerability supports the rational design of dual-metal nanotherapeutics, while also raising concerns about off-target mitochondrial injury in healthy tissues, further highlighting the necessity of tumor-specific activation.

Ferroptosis + Pyroptosis

Ferroptosis and pyroptosis cooperate to amplify anti-tumor immune responses through distinct but complementary molecular mechanisms.^75,149 Iron-dependent lipid peroxidation renders tumor cells vulnerable to oxidative damage in ferroptosis. Pyroptosis, on the other hand, drives inflammatory signaling: gasdermin proteins form membrane pores, releasing DAMPs including calreticulin, high mobility group box 1 (HMGB1), and ATP to initiate immune cascades.¹⁵⁰ Co-activation of these two immunogenic cell death (ICD) types rewrites immune-suppressive tumor microenvironments, enabling dendritic cell maturation and recruiting cytotoxic CD8⁺ T cells to tumor sites.¹⁵¹

PVP-modified NiS₂/FeS₂ hybrid nanocrystals represent a classic metal-sulfide photothermal platform.¹⁴⁹ Under near-infrared irradiation, the material generates ROS and depletes intracellular GSH, downregulating GPX4 to induce ferroptosis and activating gasdermin E to trigger pyroptosis, with pronounced abscopal effects verified in 4T1 breast tumor models.¹⁴⁹ Beyond single metal-sulfide systems, three mechanistically distinct nanoplatforms have been developed to expand the design landscape of ferroptosis-pyroptosis co-induction.

Wang’s group designed a covalent aggregation-induced emission (AIE) photosensitizer named MBTP-PA, incorporating an alkyneamide reactive warhead into its backbone.¹⁵² Inside tumor cells, the molecule undergoes thiol-yne click reactions with glutathione and cysteine, draining cellular antioxidant pools while alleviating tumor hypoxia, a critical roadblock limiting ferroptosis efficiency. Light exposure then triggers concurrent ferroptosis and pyroptosis. This metal-free design activates systemic anti-tumor immunity, erases primary tumors and suppresses metastatic lesions with mild side effects, and avoids the long-term organ metal buildup seen in sulfide nanomaterials.¹⁵²

Apart from covalent photosensitizers, iron-based nanocatalysts represent another viable design route. Feng et al constructed the HFCP iron nanomedicine to shift tumor cell death toward highly immunogenic pyroptosis rather than ferroptosis.¹⁵³ The interplay between iron carriers and loaded CP (carbonyl cyanide m-chlorophenyl hydrazone) rearranges cell death signaling, strengthening overall anti-tumor immune effects.¹⁵³ The HFCP formulation slows primary tumor growth and builds lasting systemic immune protection against untreated distant metastases. Another pH-sensitive self-assembled nanocarrier (M@P) has also been documented. Amphipathic polymers encapsulate the AIE dye MTCN-3 and TLR3 agonist Poly(I:C).¹⁵⁰ After accumulating in lysosomes, light stimulus generates reactive oxygen species and heat to rupture lysosomal membranes, turning on both ferroptosis and pyroptosis. The encapsulated Poly(I:C) works as an immune adjuvant, activating dendritic cells via TLR3 pathways and boosting IL-6 and IFN-α secretion.¹⁵⁰

While these dual cell death platforms deliver promising outcomes in preclinical animal tests, multiple translational hurdles remain unresolved. Excessive ROS production can trigger unwanted pyroptosis in healthy tissue. Without strict tumor-specific activation, gasdermin pore formation will release IL-1β and IL-18 into circulation and induce whole-body inflammation. Most metal-containing nanocarriers lack full biodegradability, and their long-term retention in vital organs creates unaddressed chronic toxic risks. Phototherapy-dependent therapies also face narrow clinical applicability, as even near-infrared window (NIR)-II light penetration is limited to only on the order of millimeters in biological tissues, restricting treatment to superficial or endoscopically accessible lesions. Moving forward, material design should center on biodegradable raw materials, tumor-restricted dual-response delivery (pH/ROS gated release, targeted GSDME regulation), and unified immune toxicity assessment standards to push these platforms closer to clinical trials.

Ferroptosis + Apoptosis

Apoptosis and ferroptosis interact closely via shared signaling nodes including p53 and mitochondrial stress signals.¹⁵⁴ The functional output of p53 varies with cellular context: it can upregulate pro-apoptotic mediators like PUMA and Bax, while simultaneously boosting expression of ferroptosis-related proteins SAT1 and GLS2.¹⁵⁵ This creates natural crosstalk linking the two cell death programs. For many tumors with impaired apoptotic signaling, ferroptosis acts as a backup killing pathway, and dual activation of both routes can effectively overcome therapeutic resistance and sustain long-term tumor suppression.^155,156

Carrier-free ASP nanoparticles target mitochondria and lysosomes to initiate caspase-mediated apoptosis. Meanwhile, they suppress GPX4 and SLC7A11 expression to trigger ferroptosis simultaneously.⁷⁹ In TNBC models, this dual therapy achieved approximately 75% tumor growth suppression in these xenograft models, outperforming either single-agent treatment.⁷⁹ These carrier-free constructs feature simple manufacturing and eliminate excipient-associated toxic risks, yet they carry an obvious drawback: no active tumor-targeting function, which weakens selective accumulation in malignant lesions.⁷⁹

The vast majority of existing dual-action nanomedicines rely on raising cellular ROS and depleting antioxidant reserves as their core working logic.¹⁵⁷ This design leaves little room to adjust the ratio of apoptotic versus ferroptotic cell death induced. In addition, the effective therapeutic range varies greatly across different tumor subtypes, depending on their inherent apoptotic competence.¹⁵⁶ Follow-up material engineering could adopt tumor-localized delivery or sequential cargo release designs. Such structures separate caspase activation and lipid peroxidation into different time windows, amplifying synergistic anti-tumor effects while reducing off-target damage to healthy tissues.

Multi-Pathway Programmed Cell Death Induction

Combining three or more cell death pathways stands as one of the most powerful bioengineering tactics to overcome broad tumor resistance. Instead of simply stacking separate cytotoxic mechanisms, this strategy leverages the cross-linked regulatory networks between different death programs. Ferroptosis breaks down cell membranes via lipid peroxidation. Pyroptosis forms inflammatory pores relying on gasdermin proteins. Apoptosis executes cell clearance through caspase cascades. Activating all three processes at once builds a self-amplifying cycle of oxidative stress and immune stimulation, an effect unattainable with single or dual-pathway nanomedicines. PANoptosis, the integrated cell death program merging pyroptosis, apoptosis and necroptosis, offers a theoretical framework to interpret the coordinated activation of multiple death signals sharing upstream regulatory factors.¹⁵⁸ This design logic carries particular value for apoptosis-deficient tumors, where ferroptosis and pyroptosis act as independent cytotoxic routes to evade common resistance mechanisms.¹⁵⁹

Ultrasound-responsive HFT nanocomposites serve as a typical triple-pathway nanoagent, assembled from hematoporphyrin monomethyl ether (HMME), ferric ions and tannic acid.¹⁶⁰ Under ultrasonic stimulation, HMME generates singlet oxygen as a sonodynamic agent. Ferric ions meanwhile consume intracellular glutathione and drive Fenton-like reactions to produce hydroxyl radicals. The combined sonodynamic and chemodynamic effects simultaneously trigger apoptosis, ferroptosis and immunogenic cell death.¹⁶⁰ In animal experiments, HFT nanoparticles combined with ultrasound remodel immunosuppressive tumor microenvironments, boost intratumoral T cell infiltration, and work synergistically with anti-PD-1 antibodies to eliminate low-immunogenic cold tumors.¹⁶⁰

Apart from HFT nanosystems, other distinct nanostructures can also initiate three coordinated cell death pathways. A self-sufficient nanozyme pPB-Fe₃O₄@CaO₂-BzATP has been developed for prostate cancer treatment.¹⁶¹ Calcium peroxide decomposes in acidic tumor microenvironments to release hydrogen peroxide and calcium ions; Fe₃O₄ nanozymes catalyze hydroxyl radical production to drive ferroptosis.¹⁶¹ Excess calcium triggers apoptotic signals, while the P2X7 agonist BzATP facilitates potassium efflux and NLRP3 inflammasome activation to induce pyroptosis.¹⁶¹ This integrated triple-death strategy achieves strong anti-tumor effects with minimal adverse reactions in both cell and animal models.¹⁶¹

Even with robust preclinical antitumor activity, triple-cell-death nanoplatforms bring extra translational challenges absent from simpler dual-action formulations. For instance, FeSA-Ir/metallene nanozymes have been shown to induce sequential ferroptosis followed by pyroptosis through light-controlled ROS production, with potent primary tumor inhibition and multi-immunogenic control of lung metastases in preclinical models.¹⁶² Nevertheless, disentangling individual pathway contributions remains inherently difficult: the relative weight of ferroptosis, pyroptosis, and apoptosis in tumor regression cannot be easily deconvolved, complicating both rational optimization and regulatory characterization.¹⁵⁹ This deconvolution problem is compounded by the reliance of most triple-responsive systems on external triggers (ultrasound, light) and multi-component architectures, which introduce batch-to-batch variability and scale-up difficulties that challenge Good Manufacturing Practice (GMP) compliance. Simultaneous engagement of multiple cytotoxic pathways also amplifies off-target risks; pyroptosis and ICD are by nature immunostimulatory, and without stringent tumor-restricted activation, systemic inflammatory complications may arise.¹⁵⁹ Our reservation is methodological: triple-death constructs currently claim success on tumor regression alone. Without pathway-resolved readouts quantifying how much death each program contributes in vivo, the added mechanisms are asserted rather than demonstrated.

Combining Ferroptosis with Conventional and Emerging Therapies

Immunotherapy and chemotherapy represent major modalities that can synergize with ferroptosis. Below, we highlight representative bioengineering strategies and unifying translational bottlenecks.

Immunogenic Ferroptosis and Its Immunomodulatory Complexity

Ferroptosis is now recognized as a distinct form of immunogenic cell death (ICD).¹¹⁰ Unlike apoptosis, which typically subdues antitumor immune responses, ferroptotic cancer cells release DAMPs including calreticulin, ATP, and HMGB1, which can stimulate antigen-presenting cells. However, ferroptosis also generates oxidized phospholipids that act in the opposite direction, directly suppressing dendritic cell (DC) function.^110,163 This intrinsic immunogenicity provides a strong rationale for combining ferroptosis inducers with immune checkpoint blockade (ICB), particularly for “cold” tumors that respond poorly to ICB monotherapy.¹⁶⁴

During ferroptosis, oxidized phospholipids further disrupt DC-mediated cross-presentation and impair CD8⁺ T cell activity. Tumor-infiltrating CD8⁺ T cells also undergo ferroptotic exhaustion by upregulating CD36, which can be reversed by CD36 blockade to restore ICB efficacy.¹¹¹ Ferroptosis also modulates macrophage polarization and NLRP3 inflammasome activity in a context-dependent manner, underscoring the need for tumor-restricted ferroptosis induction to preserve antitumor immunity.¹⁶⁵ Thus, unselective ferroptosis induction can impair antitumor immunity, supporting the need for tumor-specific triggering and T cell-protective strategies.¹¹¹

Three translational bottlenecks remain. First, systemic ferroptosis inducers threaten immune organs, as CD8⁺ T cells and DCs rely on GPX4 for redox homeostasis; immune-oriented pharmacodynamic assessments are essential for investigational new drug (IND)-enabling studies. Second, a severe pharmacokinetic–pharmacodynamic (PK–PD) mismatch exists: DAMPs release rapidly, while ICB acts with delayed kinetics, favoring stimulus-responsive sequential release systems.^20,32 Third, heterogeneous baseline expression of GPX4, SLC7A11, and ACSL4 dictates ICD responsiveness.²² 18F-FSPG PET is the most advanced candidate for non-invasive assessment of System Xc⁻ activity, though its predictive value remains to be validated. These efforts represent critical early steps, but clinical translation demands stricter focus on immune specificity, matched release kinetics, and companion diagnostics. The immune consequences of ferroptosis are thus double-edged: the same oxidized phospholipids that recruit antigen-presenting cells also blunt cross-presentation and drive CD8⁺ T-cell exhaustion. We propose that ferroptosis–immunotherapy combinations should be administered as immunomodulators rather than purely as cytotoxic agents, with timing and fractionation potentially carrying equal importance to total dose.

Bioengineering Strategies for Ferroptosis-Driven Immunotherapy

Nanoplatforms have been developed to harness ferroptotic ICD while mitigating immune suppression. One strategy uses PD-1-displaying membrane-coated nanoparticles to deliver RSL3. PD-1 sequesters programmed death-ligand 1 (PD-L1) to relieve T cell inhibition, while RSL3 induces ferroptosis, boosting CD8⁺ T cell infiltration and DC maturation in TNBC.¹²⁹ Another platform uses fucose-targeted liposomes co-loaded with 1,2-diarachidonoyl-sn-glycero-3-phosphocholine (DAPC) and CpG oligodeoxynucleotide (CpG). ROS-driven PUFA peroxidation triggers CpG release, promoting DC activation and IFNγ secretion; IFNγ further inhibits SLC7A11, amplifying ferroptosis and synergizing with anti-PD-L1 therapy.¹³⁰

Together, these platforms illustrate early engineering attempts to align ferroptotic ICD with productive antitumor immunity. Yet clinical progress remains constrained by poor immune specificity, mismatched release kinetics, and a lack of predictive companion diagnostics. These challenges demand greater focus on translational realism rather than incremental optimization of nanocarriers.

Chemotherapy

Conventional chemotherapies act by inducing ROS stress and DNA damage, and many agents can synergize with ferroptosis by overwhelming cellular antioxidant capacity and sensitizing drug-resistant tumor cells.^132,166

PSMA-targeted arsenic nanosheets (PMANs) were developed to load doxorubicin (DOX) with an exceptional capacity of 140 wt%.¹³¹ PMANs induce ferroptosis via GSH depletion and downregulation of SLC7A11 and GPX4, while DOX inflicts DNA damage and inhibits ATM to enhance chemosensitivity.¹³¹ Despite high drug loading, the clinical translation of arsenic-based nanosheets is hindered by intrinsic arsenic toxicity and unclear biodegradation pathways.

Cisplatin represents another widely studied combination partner. Liposomal co-delivery of cisplatin and the GPX4 inhibitor RSL3 has been shown to synergistically induce both apoptosis and ferroptosis in head and neck cancer models, effectively overcoming cisplatin resistance.¹³²

Despite consistent reports of preclinical synergy, several practical barriers limit clinical translation. Most chemotherapeutics lack inherent selectivity for ferroptosis, increasing the risk of exacerbated off-target damage to healthy tissues. Combining chemotherapy with ferroptosis inducers also raises the potential for overlapping toxicities, requiring careful dose optimization and scheduling. Furthermore, pharmacokinetic profiles of small-molecule chemotherapies and ferroptosis agents often differ widely. Even nanocarrier-mediated co-delivery cannot fully guarantee matched intratumoral release kinetics or optimal molar ratios for synergistic activity.

Many current platforms also bundle multiple functions without distinguishing contributions from chemotherapy, ferroptosis, or nanoparticle-induced stress. Such bundled designs hinder mechanistic understanding and complicate clinical development, echoing broader challenges of overcomplicated systems highlighted throughout this review.

Radiotherapy

Ionizing radiation (IR) generates reactive oxygen species (ROS) and induces DNA double-strand breaks, which downregulate SLC7A11 in an ATM-dependent manner, thereby enhancing ferroptosis.¹⁶⁷ This mechanism provides a rationale for combining radiotherapy with ferroptosis inducers to overcome radioresistance.

An inhalable biomimetic nanoreactor (DHA-N@M) serves as a representative radiosensitizer. It is based on polyunsaturated docosahexaenoic acid (DHA) and coated with macrophage membranes for lung cancer targeting.¹³⁴ Intratracheal administration improves tumor accumulation and reduces systemic toxicity. Under IR, DHA undergoes lipid peroxidation, while S-nitrosothiols (SNO) in the nanoreactor release nitric oxide (NO) upon glutathione (GSH) depletion. This inactivates GPX4 and amplifies oxidative stress, ultimately inducing ferroptosis and significantly suppressing orthotopic lung tumor growth (preclinical).¹³⁴

CNSI-Fe(II) has also been tested as a radiosensitizer: combined with ionizing radiation, it improved lipid peroxidation and tumor regression over either treatment alone in preclinical models.⁵⁷ Several ferroptosis-targeted nanotherapeutics have recently advanced into clinical trials. Of note, a multicenter, open-label Phase II trial of intratumoral CNSI-Fe(II) combined with radiotherapy is now recruiting in China (CTR20260473), the first formally registered clinical attempt to combine ferroptosis induction with radiotherapy.¹³³

Still, several challenges remain. Spatiotemporal coordination between radiation delivery and ferroptosis induction is critical; optimal sensitization requires precise timing of iron-based nanoparticle administration relative to irradiation. It also remains unclear whether ferroptosis inducers worsen radiation-induced normal tissue injury, and the best fractionation schedule for ferroptosis-potentiated radiotherapy has yet to be defined. Careful dose optimization and tumor-specific activation strategies are urgently needed.¹⁴⁷

Indirect Modulatory Strategies (Gas and Starvation)

Gaseous mediators including CO, H₂S, and NO modulate redox balance, GSH levels, and mitochondrial function to influence ferroptosis sensitivity.^168–170 Several nanoplatforms have incorporated gas donors to amplify ferroptosis, often alongside chemotherapy or photothermal therapy.^137,138 However, nearly all gas-based systems bundle gas release with other therapeutic modalities, making it impossible to isolate genuine ferroptosis contributions from gas signaling alone. Off-target effects on vascular function and inflammation remain poorly defined, and standalone in vivo evidence is scarce. Thus, gas-mediated ferroptosis regulation remains an exploratory strategy rather than a mature therapeutic direction.

Glucose oxidase (GOx)-mediated starvation depletes intratumoral glucose and generates H₂O₂, which can fuel Fenton chemistry and ferroptosis when paired with iron donors.¹³⁹ Representative cascade nanoreactors couple GOx and iron-based MOFs to amplify oxidative stress and trigger ferroptosis via GSH depletion and GPX4 inactivation.^139,171 Despite preclinical activity, several critical barriers hinder translation. Systemic GOx risks off-target hypoglycemia, and mismatched kinetics between H₂O₂ generation and Fe²⁺ release often limit Fenton efficiency. Moreover, GOx consumes oxygen, which can further exacerbate tumor hypoxia and impair ferroptosis. These challenges necessitate tightly coordinated, tumor-specific delivery rather than passive accumulation.

Stimuli-Responsive Systems and Theranostic Platforms

Compared with multi-pathway combinatorial therapies, two core bioengineering strategies achieve spatiotemporal control over ferroptosis. Stimuli-responsive carriers release therapeutic payloads in response to intrinsic tumor signals or external triggers, compensating for the weak tumor specificity of conventional ferroptosis modulators. Theranostic platforms, by contrast, support real-time detection of target engagement. Combined, the two strategies enable deliberate modulation of ferroptosis, rather than relying on unregulated spontaneous cell death.

Endogenous Stimuli-Responsive Systems

Endogenous responsive carriers use intrinsic tumor biochemical signals such as acidic pH, elevated ROS and hypoxia to trigger drug release without external equipment. Each cue has a characteristic failure mode. Acidification is not tumor-specific; inflamed tissues share the same pH range. ROS-responsiveness contradicts itself: ferroptosis consumes ROS, and hypoxic cores have the lowest ROS levels yet the highest ferroptosis resistance. Hypoxia offers better specificity but is unevenly distributed, and nanoparticles penetrate hypoxic zones poorly. AND-gate designs that require two cues simultaneously improve selectivity but shrink the responsive volume of the tumor. The trade-off is systematic: better specificity means less coverage, and no material innovation can resolve it without better patient selection.

pH-Responsive Systems

Acid-degradable carriers such as iron-doped ZIF-8 and CaCO₃-based nanoparticles are widely used to release Fe²⁺ in the acidic tumor microenvironment.^68,92 A representative design is the TA–Fe/ART@ZIF metal-organic framework, which co-releases Fe²⁺ and artemisinin upon acid-triggered degradation 115. For ferroptosis, the logic is direct: iron delivery should be gated by the lower pH of tumor interstitium and endosomes.^172–174 However, solid tumors display extensive pH heterogeneity. Hypoxic and necrotic regions can be neutral or even alkaline, while inflamed normal tissues share the same acidic pH range as tumors. This well-documented characteristic undermines the spatial selectivity of pH-gated delivery systems.¹⁷⁵ For iron-based ferroptosis inducers, this is not simply a matter of off-target activation—it means that a substantial fraction of the administered iron load may be released in non-tumor tissues or fail to release at all in poorly perfused tumor regions. The consequence is not uniform catalytic activity but a double failure: toxicity where it is not wanted, and inactivity where it is needed.

ROS-Responsive Systems

ROS-responsive carriers employ cleavable linkers such as thioketal, arylboronate, and peroxalate to release ferroptosis inducers upon exposure to elevated ROS in the tumor microenvironment (TME).^14,120 The conceptual appeal is clear: tumors generate higher baseline ROS than normal tissues. For ferroptosis, however, this logic contains a fundamental contradiction. Hypoxic tumor cores have the lowest ROS levels due to limited oxygen supply and often exhibit the highest ferroptosis resistance, creating conditions that directly oppose the requirements for payload release.^38,122,176 Additionally, these linkers degrade irreversibly. Any ROS generated by circulating immune cells or routine physiological metabolism will prematurely split the carrier before it reaches tumors, wasting payload and increasing off-tissue exposure.

Hypoxia-Responsive Systems

Hypoxia creates a double-edged scenario for responsive delivery. Low oxygen microenvironments stabilize HIF-1α and circHIF1α signaling to boost antioxidant activity in cancer stem cells, but they also serve as the most tissue-specific trigger available to distinguish tumors from healthy organs. Nanocarriers engineered to respond to hypoxia target areas with strong ferroptosis resistance, yet hypoxia is unevenly distributed, and standard nanoparticles struggle to penetrate deep hypoxic tumor zones.¹⁷⁷ Bacterial nanohybrids such as Ec@ZFOY have been developed to actively migrate to hypoxic regions under magnetic guidance, delivering iron-based catalysts to deep tumor regions.¹⁷⁸ Even with improved penetration, live bacterial components introduce unquantifiable immune side effects. Combined bacterial endotoxin and iron-driven ROS may trigger widespread inflammatory damage in healthy organs, and no large-animal safety data exists to validate these risks. Regulators also lack clear evaluation frameworks for bacteria-nanomaterial hybrid therapeutics, slowing their clinical translation.

AND-Gate Dual-Responsive Systems

The universal drawbacks of incomplete activation and non-selective action in single-trigger systems have spurred AND-gate designs that depend on coincident tumor-associated cues to initiate payload release. Dual-trigger AND-gate carriers theoretically reduce off-target release, yet their core working premise lacks solid validation in patient tumor samples. These formulations only release therapeutics once two distinct tumor markers appear simultaneously, for instance high ROS alongside depleted glutathione pools.¹¹⁶ SA-IO-BQR@PMEMA exemplifies this design, withholding brequinar until both signals hit functional thresholds.¹¹⁶ Perfusion shifts and cell cycle differences constantly alter microenvironmental signals, so few tumor subregions sustain all necessary triggers at once.³² This logic carries a largely untested spatial premise: although single-parameter tumor pH and hypoxia heterogeneity are well documented, quantitative co-mapping of pH, ROS, and GSH within the same human tumor microdomains has not been reported.^174,175,177 Current patient biopsy data offer little proof that stable multi-signal microdomains exist widely across clinical lesions.^32,116 This constraint stems from human tumor biology rather than imperfect nanomaterial design. For general patient cohorts, single-stimulus carriers deliver more uniform treatment responses. AND-gate platforms only show practical value for patients confirmed via imaging to carry overlapping microenvironmental signals, meaning further development of these systems must prioritize companion diagnostic screening rather than material optimization. Dual-trigger designs have not yet published in vivo tumor-to-normal selectivity data, particularly from patient-derived samples. For this reason, we advise treating AND-gate selectivity claims as elegant hypotheses rather than experimentally validated conclusions.

Exogenous Stimuli-Responsive Systems

Endogenous responsive carriers work without external equipment but inevitably compromise either tumor selectivity or intralesional coverage. Exogenous triggers including light, ultrasound and magnetic fields bypass this trade-off by manually controlling activation timing and location, though they bring separate challenges around medical hardware limits and unified clinical dosing standards.¹⁴

Light-Triggered Systems

Light stimulation delivers unmatched spatial resolution to initiate ferroptosis, with activation controllable down to millisecond timing and millimeter-scale lesion targeting.¹⁴ The chlorin e6 (Ce6) -erastin supramolecular nanoassembly stands as a typical representative: chlorin e6 photosensitizer and erastin self-assemble via hydrogen bonds and π–π stacking interactions. After illumination, Ce6 generates singlet oxygen to deplete intracellular glutathione through thiol-disulfide exchange, while erastin simultaneously blocks System Xc⁻ signaling to further amplify lipid peroxidation and ferroptosis.¹³⁵ This photodynamic-ferroptosis combined strategy relieves tumor hypoxia and strengthens anti-tumor activity, with low baseline toxicity against healthy cells observed in subcutaneous xenograft models.¹³⁵ A conceptually distinct photoinduced ferroptosis strategy involves endoplasmic reticulum (ER)-targeted probes. Xing et al developed a quinazolinone-based lipophilic probe (BODIQPy-TPA) that localizes to the ER and induces specific photoinduced ferroptosis upon light irradiation, while simultaneously enabling monitoring of ferroptosis-associated lipid dynamics through fluorescence imaging.¹⁷⁹ This design exemplifies the integration of subcellular targeting with photoresponsive ferroptosis induction.

Nevertheless, limited tissue penetration remains the principal barrier restricting clinical use. Shifting from visible light to NIR-I and NIR-II bands does improve penetration depth, yet the gap between lab results and human solid tumors stays substantial. Conventional endoscope-based optical tools only reach superficial mucosal layers; NIR-II light penetration remains limited to a few millimeters in biological tissues (eg, ~2 mm in brain), insufficient for most deep visceral tumors.¹⁸⁰

Beyond penetration depth, three practical hurdles slow clinical translation. Many clinical wards only stock 785/808 nm laser devices, while NIR-II lasers are largely confined to research labs without standardized power density guidelines or validated clinical treatment protocols. We therefore classify light-triggered ferroptosis as a locoregional modality by physics, not by choice: its realistic niche is superficial or endoscopically accessible disease, placing it alongside the Stage 1 strategies of our proposed roadmap rather than systemic ones. Meanwhile, photodynamic dosimetry lacks unified standards across studies: tissue optical properties, laser spot size and irradiation duration vary widely between papers, making cross-study comparison and human dose conversion unreliable. In addition, mainstream Type II photosensitizers require oxygen supply to produce ROS, and their activity drops sharply inside hypoxic tumor cores. Oxygen-independent Type I photosensitizers are under development but have not yet entered clinical trials.

Ultrasound-Triggered Systems

Ultrasound serves as a more clinically feasible alternative to optical excitation. Its effective penetration depth can reach up to 10 cm in soft tissues, covering most abdominal, thoracic and pelvic deep tumors.¹⁸¹ Diagnostic and therapeutic ultrasound transducers are already widely deployed in hospitals, laying a mature clinical infrastructure for sonodynamic translation.

BTO/MoS₂@CA is a typical pH-sensitive sonocatalytic nanoplatform. Few-layer molybdenum disulfide grows on tetragonal barium titanate piezoelectric substrates and is modified with cinnamaldehyde for tumor acid responsiveness. Under ultrasonic excitation, the composite gains boosted peroxidase-like activity to generate reactive oxygen species, triggering pronounced ferroptosis even inside severely hypoxic tumor areas where phototherapy fails.¹³⁶ Ultrasound-responsive microbubbles are another promising route: ultrasonic cavitation generates mechanical force to temporarily open biological barriers, disrupt tumor cell structure and boost intracellular oxidative stress to drive lipid peroxidation and ferroptosis.¹⁸¹ Biomimetic nanocarriers activated by ultrasound can also break intracellular redox balance to induce ferroptotic cell death.¹⁸²

Several unresolved practical issues block unified clinical sonodynamic standards. Ultrasound parameters lack consistent calibration across studies—frequency, duty cycle, acoustic pressure and irradiation duration differ drastically between experimental designs.¹⁸¹ No unified sonodynamic dose metric equivalent to radiotherapy’s Gray unit has been established, which regulators view as a major obstacle for IND filing. Cavitation, sonoluminescence and piezocatalysis together create complex multi-layer mechanisms hard to isolate and quantify.^136,181 Inertial cavitation brings non-specific damage: mechanical shockwaves induced by ultrasound can trigger multiple forms of cell death simultaneously, making it difficult to separate ferroptosis-specific therapeutic effects from unrelated cytotoxicity.¹⁸¹ Commercial ultrasound devices from different manufacturers carry inconsistent waveform output and focal zone parameters, and none come with pre-set sonodynamic therapy modes.¹⁸¹ Though microbubble co-delivery can partially ease these issues, it introduces new risks around formulation stability and batch consistency.¹⁸³

Magnetic-Triggered Systems

Magnetic stimulation has a unique advantage over light and ultrasound: magnetic fields pass through human tissues without attenuation, removing tissue depth limitations entirely. Magnetic hyperthermia equipment such as MagForce has already obtained European approval for glioma treatment, forming a ready clinical foundation for ferroptosis-related magnetic platforms.¹⁸⁴

Iron oxide nanozymes represent the core magnetic therapeutic carriers. When exposed to alternating magnetic fields (AMF), they generate localized thermal energy to upregulate ACSL4 and accelerate lipid peroxidation, thereby inducing ferroptosis.¹⁸⁵ Still, three interconnected practical obstacles delay clinical transformation. Spatial accuracy is poor: AMF heating zones span centimeter scale, so thermal damage cannot be fully confined to tumor tissues adjacent to vital organs like the brain and major blood vessels. Eddy currents also heat surrounding muscle and skin tissue, forcing researchers to use weaker magnetic field strengths and limiting the temperature rise needed to trigger sufficient ferroptosis. Long-term biosafety risks cannot be ignored either: injected iron oxide nanoparticles accumulate in liver and spleen for weeks to months, and persistent iron overload may trigger off-target lipid peroxidation in normal organs.¹⁸⁵ Besides, most preclinical AMF parameter settings mismatch approved clinical devices, complicating dose extrapolation to human subjects. Pulsed electromagnetic fields and magneto-mechanical designs are being explored as alternative routes, yet all remain at early preclinical proof-of-concept stages. Material optimization alone cannot resolve heating precision, device standardization and large-animal chronic toxicology issues, which are core bottlenecks for clinical progress.¹⁸⁵ Of the exogenous stimuli-responsive platforms surveyed in Table 3, magnetic field-triggered systems (MFCF) have reached the highest reported evidence level, as demonstrated by studies using orthotopic immunocompetent models. Even so, they still face the previously enumerated barriers of heating precision and device standardization, underscoring that evidence tier alone does not guarantee clinical readiness.

Table 3.

Comparison of Endogenous and Exogenous Stimuli-Responsive Systems for Ferroptosis-Targeted Therapy

Stimulus Type	Representative Systems	Core Advantages	Key Challenges/Translational Limitations	Highest Evidence Level (Model Tier)	Translational Readiness
pH (endogenous)	Acid-degradable MOFs (TA–Fe/ART@ZIF) (Fe2⁺ + artemisinin);114 CaCO3-based nanoparticles68,92	Autonomous activation without external energy input	Poor tumor specificity (acidic inflamed and gastric tissues); heterogeneous intratumoral pH distribution	Preclinical murine models	Low
ROS (endogenous)	ROS-cleavable linkers (thioketal, arylboronate, peroxalate) (ferroptosis-inducer payloads, generic)120	Higher baseline specificity than pH-responsive counterparts	Low ROS levels in hypoxic cores; ROS self-consumption during ferroptosis; off-target activation in inflamed tissues	Preclinical murine models	Low
Hypoxia (endogenous)	Bacterial nanohybrids (Ec@ZFOY) (iron-based catalysts)178	Active penetration into deep hypoxic tumor zones	Biosafety risks of live bacteria; immune reactivity; strict GMP barriers; incomplete intratumoral coverage	Preclinical murine models	Low
Light (exogenous)	Ce6–erastin supramolecular nanoassembly (erastin)135	Millimeter-scale spatial and millisecond temporal precision	Limited tissue penetration (<1 cm); restricted to superficial or endoscopically accessible tumors	Preclinical murine models	Low (niche use for superficial lesions)
Ultrasound (exogenous)	pH-sensitive sonocatalyst (BTO/MoS2@CA) (catalytic nanozyme; no diffusible payload)136	Deep tissue penetration (>10 cm); non-ionizing modality	Lack of standardized clinical dosimetry; non-specific cavitation injury; ambiguous cell-death mechanisms	Preclinical murine models	Low
Magnetic field (exogenous)	Dual-magnetic nanozyme (MFCF) (catalytic nanozyme; no diffusible payload)186	Unlimited penetration depth; reduced EPR dependence; MRI-guidable accumulation	Off-target eddy-current heating; low spatial resolution of heating control; long-term metal retention	Orthotopic glioblastoma model, immunocompetent mice	Moderate
AND-gate multi-responsive	SA-IO-BQR@PMEMA (brequinar) (ROS + GSH dual activation)116	Ultra-high tumor specificity via logical dual-signal activation	Mismatched stimulus thresholds in heterogeneous tumors; incomplete intratumoral activation; complex fabrication; no validated multiplexed imaging readout	Preclinical murine models	Low

Note: Translational readiness is assigned by an objective evidence-based rubric, not qualitative judgment: High = registered clinical trial or published clinical data; Moderate = efficacy demonstrated in orthotopic or immunocompetent (syngeneic) in vivo tumor models; Low = efficacy demonstrated only in subcutaneous xenograft or immunodeficient models; Concept validation = in vitro evidence only. Evidence levels denote the highest model tier reported for each representative system in the cited primary studies. The clinically most advanced ferroptosis nanomedicine, intratumoral CNSI-Fe(II), is a locoregionally delivered Level 1 platform rather than a stimuli-responsive system and is therefore not listed in this table. No stimuli-responsive ferroptosis system in this table has yet entered clinical evaluation.

Abbreviations: ART, artemisinin; CaCO₃, calcium carbonate; Ce6, chlorin e6; CNSI-Fe(II), carbon nanoparticle–iron(II) complex; EPR, enhanced permeability and retention; GMP, good manufacturing practice; GSH, glutathione; MOF, metal-organic framework; MRI, magnetic resonance imaging; ROS, reactive oxygen species; TA, tannic acid; ZIF, zeolitic imidazolate framework.

Hybrid Trigger Systems

Each single physical excitation mode carries intrinsic limitations, so researchers have constructed multi-modal hybrid nanoplatforms to offset the respective flaws of individual stimulation types, with multiple designs already verified in preclinical trials.¹⁸⁴ One typical design is iron-based magnetic nanocomplex Fe₃O₄-SRF@FeTA.¹⁸⁷ Its magnetic core enables tumor-specific accumulation and photothermal conversion, while the ferric tannate shell responds to pH and light signals to release loaded sorafenib, a classic ferroptosis inducer.¹⁸⁷ After magnetic-guided enrichment at tumor sites, subsequent laser irradiation activates dual photothermal and chemodynamic pathways to boost ferroptosis, yielding substantial anti-tumor outcomes in animal models.¹⁸⁷

Dual magnetic-responsive nanozyme MFCF represents another viable candidate. Functionalized with folic acid for tumor molecular targeting, it reacts to both static and alternating magnetic fields. When exposed to AMF and near-infrared light simultaneously, the material generates large quantities of reactive oxygen species and consumes intracellular glutathione; its synergistic magnetothermal and type-II photodynamic effects also relieve tumor hypoxia. This cascade finally triggers ferroptosis, apoptosis and immunogenic cell death, alongside pronounced intratumoral T cell infiltration.¹⁸⁶

Magnetic microbubbles paired with ICG-carrying liposomes form another hybrid therapeutic system. Magnetic navigation improves nanoparticle tumor enrichment, and ultrasound cavitation further boosts local permeability. The encapsulated ICG absorbs 808 nm laser energy to produce mild heat and photodynamic reactions. Elevated temperature and excess ROS together deplete glutathione pools and initiate ferroptosis, which in turn reinforces photodynamic treatment efficacy through positive feedback.¹⁸⁸ Porphyrin-modified Zr-based MOF loaded with zinc ferrite (named PZFH) expands the library of multi-stimulus nanozymes. This composite delivers sequential oxidase activity under light irradiation and magnet-enhanced peroxidase capacity, working together to strengthen ferroptotic damage in tumor cells.¹⁸⁹

Theoretically, such hybrid platforms take advantage of magnetic guidance to gather nanocarriers inside deep tumor tissue, followed by light or ultrasonic stimulation to locally trigger ferroptosis on demand.¹⁸⁴ Yet, multiple core questions still lack systematic experimental validation. For instance, there is no clear consensus on the optimal sequence between magnetic enrichment and optical/ultrasonic excitation. Researchers also have not fully clarified how overlapping physical stimuli may damage normal surrounding tissue, nor quantitatively distinguished genuine synergistic effects from simple additive cytotoxicity.¹⁸⁴ Though exogenous responsive nanomaterials show precise controllability under laboratory settings, their clinical translation hinges more on standardized medical hardware and unified therapeutic protocols, rather than minor structural modifications to nanocarriers themselves. Each additional external modality multiplies device dependencies, dosimetric variables and regulatory touchpoints; we question whether hybrid triggers can clear the reproducibility thresholds that single-modality systems already struggle to meet. The distinctions in target selectivity, tissue penetration and clinical maturity between endogenous and exogenous responsive ferroptosis nanotherapies are systematically laid out in Table 3.

Theranostic Platforms

Theranostic nanocarriers combine therapeutic payloads with imaging moieties to validate target engagement in vivo, yet this dual design brings inherent trade-offs. Attaching imaging labels inevitably cuts maximum drug loading capacity, enlarges particle hydrodynamic size, and complicates batch manufacturing workflows—all while conventional computed tomography (CT) and MRI devices already support routine clinical lesion assessment without custom nanoscale tracers.

Imaging probes designed to track ferroptosis can be categorized based on the actual biological signals they capture, rather than broad functional labels commonly used in literature. Five distinct classes cover sequential events across the ferroptosis cascade: lipid peroxidation, labile iron pools, System Xc⁻ transporter activity, intracellular glutathione concentration, and general reactive oxygen species accumulation. Each marker reflects a separate step ranging from upstream metabolic transport to downstream lethal membrane damage.^14,122 Still, no single imaging agent can independently confirm full ferroptotic cell death.¹⁴ Lipid peroxides may build up even when cells retain functional antioxidant defenses and avoid death; elevated labile iron is a necessary yet insufficient trigger for Fenton reactions; System Xc⁻ signals only reflect cystine uptake rather than cell fate; shifts in GSH or ROS levels also arise from dozens of unrelated cellular stress pathways.^32,38,122 The following sections unpack each probe category, clearly defining the biological signals they detect and the interpretive limits researchers must bear in mind during data analysis.

Lipid Peroxidation Probes

C11-BODIPY 581/591 serves as a classic fluorescent reporter for lipid peroxidation in cell culture systems, yet its short-wavelength emission severely restricts deep tissue visualization. More critically, the probe cannot distinguish lipid peroxidation induced by ferroptosis from oxidative damage triggered by other cell death pathways.¹⁹⁰ While lipid peroxidation marks the terminal execution stage of ferroptosis, elevated peroxide levels do not equate to irreversible cell death.¹²² Cells with intact GPX4 or FSP1 antioxidant function can clear accumulated lipid peroxides and survive, even with positive C11-BODIPY signals.^34,38 Beyond conventional lipid peroxidation sensors, ER-targeted probes have been developed that integrate ferroptosis induction with real-time lipid dynamic monitoring.¹⁷⁹ A positive fluorescent readout only verifies lipid oxidative stress, and cannot be treated as definitive proof that cells will undergo ferroptosis.

System Xc⁻ Activity Probes

¹⁸F-FSPG PET enables non-invasive quantification of System Xc⁻ transport activity and has been evaluated in clinical trials for head and neck cancer, colorectal cancer, and non-Hodgkin lymphoma, as well as prostate cancer.^176,191 These early clinical data have established the tracer’s safety and biodistribution profile in humans, but its predictive value for ferroptosis-directed therapy remains entirely hypothetical: no prospective trial has paired baseline FSPG uptake with subsequent ferroptosis inducer treatment to determine whether uptake correlates with therapeutic response, and it has not obtained regulatory approval as a companion diagnostic for this indication. Practical barriers also slow widespread clinical adoption: the radiotracer has a 110-minute half-life that demands on-site radiochemistry production, and uniform cut-off SUV thresholds have not been validated across tumor subtypes.¹⁹¹ Beyond logistical hurdles, its core functional limitation lies in narrow pathway specificity. It tracks System Xc⁻–mediated cystine transport as an upstream defense activity, but cannot report cell death execution triggered by GPX4 inhibition, iron overload, or FSP1/DHODH blockade. In short, ¹⁸F-FSPG PET reflects a specific metabolic vulnerability rather than a universal ferroptosis biomarker; its clinical utility depends on prospective validation in SLC7A11-targeted regimens, and it is unlikely to serve as a pan-ferroptosis companion diagnostic.¹⁹¹

Labile Iron Probes

Art-Gd MRI contrast agents capture intracellular labile iron buildup and have shown preliminary performance in preclinical models of cardiac and renal ferroptotic injury, yet imaging relies on high-field 7-T scanners; its diagnostic reliability on standard clinical 1.5T and 3T equipment remains unvalidated.¹⁹² Elevated free iron represents a mandatory prerequisite for Fenton-mediated lipid peroxidation, but iron overload alone does not push cells toward death if downstream antioxidant pathways remain intact.³² Multiple small-molecule fluorescent sensors have been developed to detect Fe²⁺ and Fe³⁺ in cell culture, yet poor tissue penetration and rapid systemic clearance severely hinder their in vivo application.^193–195

Glutathione and ROS Probes

A wide array of fluorescent sensors such as monochlorobimane, RealThiol, DCFH-DA and dihydroethidium are available to trace intracellular glutathione and reactive oxygen shifts within cultured cells.^196,197 Still, these markers capture universal stress signals rather than ferroptosis-specific events. GSH depletion occurs under numerous metabolic disturbances, and ROS bursts accompany apoptosis, necroptosis and pyroptosis alike.^32,122 None of these tracers can isolate ferroptotic oxidative damage from other regulated cell death forms, and none support non-invasive whole-animal imaging. Their use is therefore limited to supplementary readouts under tightly controlled in vitro experimental setups, and cannot act as standalone pharmacodynamic biomarkers.

Beyond ¹⁸F-FSPG PET and Art-Gd MRI, dozens of organelle-targeted fluorescent probes have been engineered to track ferroptosis-associated biochemical changes in cell lines.^198–200 There exists an inverse correlation between the abundance of these lab-based tools and their clinical readiness. Almost all such probes are restricted to cellular assays, held back by three shared limitations: insufficient pathway selectivity (C11-BODIPY reports generic lipid oxidation instead of ferroptosis-specific damage), delayed signal generation that only captures late-stage peroxidation, and unfavorable in vivo properties including weak tissue penetration and rapid off-target clearance.^14,190 This disparity between rich in vitro tool development and scarce clinical tracers reinforces a core theme of this review: the design of novel detection probes does not equal meaningful translational progress.

Given all these technical constraints, imaging’s most realistic clinical role falls under open-loop guidance, rather than fully closed-loop real-time therapeutic modulation.¹⁹¹ Under this framework, pretreatment scans such as ¹⁸F-FSPG PET first filter patient subgroups with high System Xc⁻ activity who stand to benefit from cystine transport inhibitors. Patients then receive integrated theranostic nanocarriers carrying ferroptosis inducers paired with imaging reporters like Art-Gd. Post-treatment imaging can then confirm intratumoral iron release and lipid peroxidation to validate target engagement, guiding adjustments to subsequent treatment cycles. This workflow does not rely on instantaneous dose re-dosing during scanning sessions; imaging functions for pre-screening and post-treatment response verification only.

The theoretical concept of closed-loop theranostics, where imaging readouts trigger immediate supplementary drug delivery during one hospital visit, has been proposed by analogy to glucose-responsive insulin carriers.¹⁷ This design, however, mismatches real clinical workflows. Ferroptotic damage develops over several hours, while PET and MRI scans generate static snapshot data with lengthy acquisition windows.²⁰¹ By the time imaging results are available for dose adjustment, the optimal therapeutic window has already passed. Additionally, keeping patients inside scanners for repeated drug administration brings heavy time and cost burdens. Closed-loop therapeutic control remains a long-term conceptual goal, rather than an immediate priority for nanomaterial engineering. For the near term, open-loop imaging guidance represents the most pragmatic route to incorporate ferroptosis monitoring into routine oncology practice.

Critical Appraisal and Future Horizons: From Technical Success to Clinical Reality

The preceding chapters have outlined the rapidly evolving bioengineering toolbox for ferroptosis therapy, including targeted delivery, multi-pathway combination, smart responsiveness, and theranostic integration. Yet clinical translation remains stalled not for lack of molecular mechanistic insight, but because of persistent mismatches between preclinical design and human pathophysiology. This chapter moves beyond technical description to confront core biological and translational realities, with a focus on model validity, clinical predictivity, and pragmatic paths forward.

The “Mouse vs Human” Reality Gap

The vast majority of ferroptosis nanoplatforms are evaluated in subcutaneous xenografts grown in young, immunodeficient mice.⁸⁸ These models are convenient for screening but systematically overestimate therapeutic efficacy. Human solid tumors contain dense stromal matrices, high interstitial fluid pressure, and heterogeneous perfusion that severely restrict nanoparticle penetration and retention.²⁰¹ Immunodeficient hosts lack adaptive immune networks including CD8⁺ T cells and IFNγ signaling, which directly suppress SLC7A11 and modulate ferroptosis sensitivity in vivo.²⁰² Studies in orthotopic tumors and patient-derived models further expose the limitations of subcutaneous xenografts: the doses required to overwhelm liver sequestration in mice do not translate to human patients, and the homogeneous, immunocompromised microenvironment of subcutaneous models bears little resemblance to clinical solid tumors. Orthotopic models, patient-derived xenografts, and humanized immune systems better recapitulate clinical physiology but remain drastically underused.^112,203,204

This mismatch between preclinical models and human tumor biology reflects a field-wide flaw in experimental design, rather than a limitation of material science, and it undermines the predictive value of most published efficacy data.

Against this backdrop of flawed models, it is relevant to examine what has been tested clinically in humans. To date, only one ferroptosis-directed nanomedicine has entered clinical evaluation. CNSI-Fe(II) has completed a Phase I trial in 19 patients with advanced solid tumors in China (NCT06048367);¹¹⁵ a Phase Ib/IIa trial is registered under NCT07433283;¹¹⁷ a multicenter, open-label Phase II trial combining intratumoral CNSI-Fe(II) with radiotherapy is recruiting in China (CTR20260473).¹³³ It is important to recognize the limited scope of this evidence: the agent is administered locoregionally, its regulatory approval is specific to China, the published data are limited to a small cohort from a single country, and the indication is strictly confined to accessible solid tumors amenable to intratumoral injection. Within these narrow boundaries, the trial reported that CNSI-Fe(II) consistently induced marked central necrosis on imaging, yet this response was poorly captured by conventional Response Evaluation Criteria in Solid Tumors (RECIST) criteria. Four of 19 patients (21.1%) with heavily pretreated, progressive disease achieved long-term survival of 28.9–36.2 months and regained sensitivity to previously failed systemic therapies. These observations provide early human evidence that ferroptosis induction is achievable and can resensitize resistant tumors in a locoregional setting, but they should not be extrapolated to systemic ferroptosis nanomedicine or to broader patient populations. The data also highlight the critical need for ferroptosis-specific response metrics. It should be noted, however, that the observed tumor necrosis has not been confirmed by biopsy-based pharmacodynamic markers (eg, GPX4 loss, lipid peroxidation products) to be ferroptosis-specific; physical embolism and local inflammation may contribute to the imaging findings. Thus, while CNSI-Fe(II) is mechanistically defined as a ferroptosis inducer in preclinical models, its mode of action in patients remains to be rigorously validated.¹⁰² This also highlights the evidentiary gap: while CNSI-Fe(II) represents the sole platform with published clinical data, all other combination strategies in Table 2 remain at preclinical evidence levels, underscoring the field’s reliance on non-human models for most proposed regimens.

The limited clinical pipeline beyond CNSI-Fe(II) is itself revealing. Brequinar, a DHODH inhibitor with strong preclinical ferroptosis activity, was evaluated in a Phase Ib/II trial for relapsed/refractory acute myeloid leukemia (AML) (NCT03760666) but terminated due to lack of efficacy.²⁰⁵ Auranofin, a gold compound that increases ROS, had a Phase I trial in advanced solid tumors withdrawn (NCT02126527).²⁰⁶ These two failures, one due to lack of efficacy and the other for undisclosed reasons, highlight a recurring pattern. Preclinical ferroptosis activity does not guarantee successful clinical translation, and the absence of predictive biomarkers means these trials have essentially proceeded with blind dosing.

Collectively, these clinical data indicate that ferroptosis can be induced in human tumors, but the metrics we use to measure it (RECIST) and the targets we prioritize (single-node inhibition) are misaligned with clinical reality. The model limitations discussed above are not abstract concerns. They carry tangible implications for trial design, outcome assessment, and data interpretation. The following sections examine how these misalignments manifest in specific translational assumptions: the safety of non-degradable metal platforms and the viability of biomarker-blind combination strategies. Across the broader ferroptosis clinical pipeline, three additional small-molecule agents, namely eprenetapopt, withaferin A, and altretamine, have been evaluated in Phase I/II settings, though their contributions to ferroptosis are less well defined mechanistically or have not been isolated in clinical practice.²⁰⁷ None has advanced beyond early-stage testing or provided evidence of tumor-specific ferroptosis in patients. The clinical landscape remains disproportionately thin, and CNSI-Fe(II) stands alone as the only nanomedicine platform with human efficacy data.^206,208

This translational framework aligns closely with the roadmap recently proposed by Kang et al in Nature Reviews Clinical Oncology, which similarly emphasizes biomarker-guided patient selection, rational combination strategies, and the critical gap between preclinical models and human tumor physiology as the central hurdles to clinical progress.²⁰⁷

The EPR Effect in Human Tumors: Reassessing Passive Targeting

The enhanced permeability and retention (EPR) effect serves as the default rationale for passive tumor targeting across nearly all ferroptosis nanoplatforms. The concept is elegant: leaky tumor vasculature and impaired lymphatic drainage should allow nanocarriers to preferentially accumulate in malignant tissue. But elegance in theory does not guarantee reliability in patients.

A landmark meta-analysis by Wilhelm et al found that only a median of 0.7% of systemically administered nanoparticles accumulate in solid tumor tissue, a concentration an order of magnitude below the threshold required for consistent on-target efficacy.⁶⁹ This 2016 estimate has since been contextualized by emerging human imaging data. A 2024 PET imaging study using ⁸⁹Zr-labeled starPEG nanocarriers in prostate cancer patients revealed that EPR-mediated tumor uptake is highly phenotype-dependent, ranging from high accumulation in EPR-permissive models to negligible uptake in others, with heterogeneous peripheral distribution rather than uniform tumor penetration.²⁰⁹ Complementary reviews published in Nature Reviews Bioengineering in 2024 have further challenged the EPR paradigm, proposing that nanoparticle entry into tumors is better explained by active endothelial transport processes known as the active transport and retention (ATR) principle, rather than passive extravasation alone.^21,210 This evolving mechanistic understanding reinforces that measurable EPR-driven accumulation occurs in fewer than 20% of human solid tumors and shows no consistent correlation with objective treatment response. This is not a limitation that better nanomaterial design can fix; it reflects the fundamental physiology of human tumors, which maintain active lymphatic drainage and endothelial transport regulation that murine xenografts do not recapitulate.

The consequences are not theoretical. Multiple late-stage nanoparticle candidates, including several transferrin-targeted iron-oxide platforms, were terminated in Phase II trials not for systemic toxicity, but because intratumoral drug exposure fell below the threshold required for measurable efficacy. The EPR effect, reliable and predictable in subcutaneous murine models, does not translate into a dependable delivery mechanism for most human cancers.

The implications are particularly pronounced for ferroptosis nanomedicines. Induction of ferroptosis relies on catalytic iron and sustained oxidative stress, both of which demand a threshold concentration of the inducer within tumor tissue. Subthreshold accumulation does not deliver partial efficacy; it yields no therapeutic benefit at all, while still exposing normal tissues to potential toxicity. Platforms that rely exclusively on passive extravasation, with no active targeting, AND-gate responsivity, or locoregional administration, therefore lack clinical viability. This limitation arises not from subpar formulation design, but from a fundamental mismatch between their underlying delivery paradigm and the physiology of human tumors.

The dose threshold problem compounds this delivery failure. Ouyang et al showed that doses above approximately 1 trillion nanoparticles per mouse can overwhelm Kupffer cell clearance and substantially improve tumor delivery, with reported values reaching up to 12% of the injected dose per gram under optimized conditions.¹¹² This threshold, however, is unlikely to be clinically achievable in humans. If the delivery mechanism fails to function at clinically achievable doses, the catalytic potency of any ferroptosis inducer becomes irrelevant. This principle applies to the entire paradigm of passive tumor targeting.

This failure is not a limitation of individual carriers; it is a shared vulnerability across the entire passive-targeting paradigm, in which materials are chosen for their catalytic potency rather than their translational viability. If the delivery mechanism does not work at clinically achievable doses, the catalytic potency is irrelevant.

Functional Overdesign: The Hidden Cost of Complexity

This observation extends well beyond material choice to shape the underlying logic of platform design. Across published work, nanocarriers are routinely built with accumulating layers of function: responsive gating to multiple cues, imaging moieties, combined payloads, and active targeting ligands. The FeSHS platform described earlier illustrates this trajectory. It pairs three enzyme-like activities, DHODH inhibition, and iron-doped silica within a single construct, yet its in vivo efficacy does not differ meaningfully from much simpler iron-based formulations.

Published work rarely includes head-to-head comparisons between single-stimulus and multi-functional ferroptosis platforms. This gap reflects a wider field preference for functional novelty over rigorous side-by-side validation. The magneto-photo-acoustic MEPNP platform follows a similar trajectory, combining magnetic targeting, near-infrared-triggered erastin release, and multimodal imaging in one system. It achieves marked antitumor effects in TNBC models, including primary growth inhibition, reduced lung metastasis, and median survival exceeding 45 days, but the study provides no direct comparison of the full integrated platform against its individual functional components. The 6198 differentially expressed genes detected after treatment further highlight how mechanistic attribution grows increasingly ambiguous as additional functions are layered onto a single carrier.²¹¹

Each added function increases batch-to-batch variability, quality-control difficulty, and regulatory risk. Many “smart” designs that are celebrated in academic journals become liabilities in clinical translation, where the core requirements are scalability, reproducibility, and well-defined safety profiles. This does not mean complexity is never useful, but it should be justified by a defined clinical need rather than by the logic of “more functions are better.” These clinically uncertain directions are symptoms of the same underlying problem: a field that has prioritized functional demonstration over translational discipline.

Reorienting the Field: A Tiered Roadmap from Local Proof-of-Concept to Systemic Therapy

The four-level engineering framework evaluated above classifies platforms by functional integration; it is not a ranking of clinical promise, and the path below is not a march from Level 1 to Level 4. If anything, the order runs in reverse. Progress will not come from pursuing all strategies simultaneously, but from advancing them in order, using each tier’s clinical data to de-risk the next. Three sequential stages structure this trajectory.

Stage 1: establishment of a clinical foothold through Level 1 platforms delivered locoregionally. The fastest path to human proof-of-mechanism does not involve systemic administration at all. Intratumoral injection, inhalable formulations, and biodegradable hydrogel depots bypass both the EPR constraint and systemic on-target toxicity entirely. Among Level 1 platforms, CNSI-Fe(II) is the clear lead candidate, with validated manufacturing and early clinical safety data from a first-in-human cohort.¹⁰² Crucially, this prioritization must be understood with its full clinical context: intratumoral injection is a marginal palliative route in routine oncology practice, confined to superficial or endoscopically accessible lesions in patients with limited options, and unsuitable for deep-seated tumors or metastatic disease. The primary value of this cohort is therefore not therapeutic efficacy in localized disease, but the first human pharmacokinetic, iron-release, and on-target toxicity data that can calibrate dosing and safety parameters for all subsequent systemic designs. Locoregional delivery is a research tool to generate human mechanism data, not a clinical solution to be scaled. The field has historically pursued systemic formulations before establishing whether ferroptosis can be safely induced in humans at all; locoregional administration corrects this inversion by providing mechanism-of-action data in the least confounded manner. Once those data are available, the translational path must rapidly transition to systemic platforms capable of treating metastatic disease, which remains the dominant cause of cancer mortality. Locoregional delivery is the entry point, not the destination; its role is to provide the safety and target-engagement benchmarks that make systemic trials interpretable, which is precisely why it is the first, not the final, step in the roadmap.

Stage 2: prospective clinical qualification of companion diagnostics, particularly the imaging components that enable Level 4 theranostic integration, to unlock Level 2 combination systemic therapy. Once locoregional data confirm that ferroptosis induction is achievable and tolerable in humans, the rate-limiting step for systemic therapy shifts from delivery to patient selection. Among the candidate stratification tools, ¹⁸F-FSPG PET offers the strongest rationale for SLC7A11-targeted regimens. However, its predictive value remains hypothetical until validated in prospective trials that pair baseline uptake with subsequent ferroptosis inducer treatment and correlate SUVmax with clinical outcome. By contrast, for GPX4-directed or iron-based ferroptosis inducers, FSPG PET is at best an indirect marker of pathway vulnerability rather than a direct pharmacodynamic readout; response may depend more on GPX4 expression, ACSL4 activity, or the labile iron pool, requiring complementary biomarker panels for those regimens. Formal validation of ¹⁸F-FSPG PET, or any other candidate biomarker, as a predictive companion diagnostic will serve as the mandatory bridge between local proof-of-concept findings and enrichment-designed systemic trials. Without such validation, even well-formulated combination systems will fail in unselected patient cohorts, as heterogeneous baseline pathway activity guarantees a large non-responder population.

Stage 3: standardized GLP toxicology for iron-retaining nanomaterials as the gate for Level 3 stimuli-responsive systemic platforms. Level 3 responsive systems offer the highest theoretical precision, but they cannot enter meaningful clinical development until two prerequisites are met. First, a standardized 90-day non-rodent GLP toxicology framework for iron-retaining nanomaterials must be established, starting with the clinically most advanced candidate, CNSI-Fe(II).^102,212

What distinguishes ferroptosis-directed GLP toxicology from conventional safety evaluation is the need for iron-specific and ferroptosis-specific endpoints: conventional markers of organ injury (alanine aminotransferase/aspartate aminotransferase) may not capture the early oxidative damage that precedes overt necrosis. These markers should be treated as exploratory pharmacodynamic endpoints in early-phase trials rather than validated registration-grade toxicity metrics. Ferroptosis-specific PD markers that should be prioritized include lowered GPX4 activity, elevated malondialdehyde and 4-hydroxynonenal (4-HNE), depleted glutathione, and expanded labile iron pools.^201,213,214 Their limitations are also real: GPX4 activity and GSH are ex vivo labile and require rapid sample processing (eg, NEM alkylation for GSH); serum malondialdehyde (MDA) and 4-HNE are non-specific oxidative stress markers and should be interpreted as a panel rather than as standalone ferroptosis evidence; LIP is best measured in biopsy tissue by ICP-MS rather than in circulation. These markers should be measured alongside iron-specific endpoints to establish a complete ferroptosis pharmacodynamic profile in early-phase trials.

For the kidney, urinary kidney injury molecule 1/neutrophil gelatinase-associated lipocalin/N-acetyl-β-D-glucosaminidase should be included as sentinel markers of proximal tubular injury—they rise earlier than traditional serum markers and are mechanistically aligned with the site of ferroptotic damage.^215,216 For the liver, the critical metric is not transaminase elevation but quantitative iron burden; MRI R2 mapping or biopsy-based inductively coupled plasma (ICP)- mass spectrometry (MS) should be used to establish iron accumulation kinetics as a dose-limiting parameter.^40,216 And because ferroptosis generates lipid peroxidation products that persist beyond the acute treatment window, serum and urinary 4-HNE, 8-iso-PGF2α, and malondialdehyde should be monitored longitudinally to assess chronic oxidative stress between dosing cycles. The second prerequisite is the qualification of companion diagnostic endpoints to verify intratumoral trigger fidelity. Only once these safety standards and verification tools exist will complexity add clinical value rather than regulatory risk. The US Food and Drug Administration’s final guidance “Drug Products, Including Biological Products, that Contain Nanomaterials” (issued April 2022) adopts a risk-based, case-by-case evaluation framework, requiring comprehensive characterization of particle size, surface charge, drug release kinetics, stability, and degradability.²¹⁷

For the coming decade, the field’s strategic mistake has been to pursue the most sophisticated designs first. The viable path runs in the opposite direction: start with the simplest platforms delivered locally, use biomarkers to expand to systemic combinations, and only then add layers of responsive control. Ferroptosis therapy will advance not through technical complexity, but through sequential, data-driven translational discipline.

Conclusions

Research targeting ferroptosis as a cancer treatment has laid solid mechanistic groundwork, yet clinical evidence remains confined to early-stage outcomes of one locally injectable nanomedicine, CNSI-Fe(II). The primary roadblocks are pharmacological and clinical rather than material-design flaws: off-target risks driven by the ubiquitous expression of ferroptosis regulators in healthy organs, poor reproducibility of the EPR effect in human tumors, and the absence of standardized long-term toxicology frameworks for metal nanocarriers. A stepwise path forward is therefore indicated: local delivery vehicles to generate human proof-of-mechanism; prospective validation of imaging biomarkers such as ¹⁸F-FSPG PET to stratify patients; and standardized 90-day GLP toxicology focused on iron-specific endpoints before systemic responsive platforms can be developed. These conclusions are tempered by evidentiary limitations: most preclinical data derive from cell cultures and subcutaneous immunodeficient models, with safety evaluations rarely extending beyond acute timeframes. We argue that the next decade of ferroptosis nanomedicine will be decided less by materials innovation than by pharmacological discipline: iron-aware enrollment criteria, imaging-verified target engagement, and long-term toxicology standards that do not yet exist. Platforms that ignore these constraints, however elegant their chemistry, are unlikely to reach patients.

Funding Statement

No funding was received for this review article.

Data Sharing Statement

Data sharing is not applicable to this article as no new data were created or analyzed in this study.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors report no conflicts of interest in this work.