Autophagy is a highly organized catabolic process essential for cellular homeostasis. Its initiation is triggered by stressors such as nutrient deprivation or oxidative stress, which activate the unc-51 like kinase 1 (ULK1) complex. ULK1 directly interacts with lactate dehydrogenase (LDH)A, phosphorylates serine-196 and elevates intracellular lactate levels. This promotes lactylation of vacuolar protein sorting 34 (VPS34), which in turn amplifies autophagic flux and endolysosomal trafficking (15). VPS34, a class III PI3K, produces phosphatidylinositol-3-phosphate to recruit autophagy-related (ATG) proteins and initiate autophagosome nucleation (16). This cascade is not merely a housekeeping pathway but a dynamic stress response. For example, in KRAS^G12C-driven LC cells, KRAS^G12C inhibition induces ULK1/2-mediated autophagy as a critical survival mechanism, highlighting its potential as a pharmacological target (17). However, this dependency is likely not universal and a systematic understanding of how different oncogenic drivers rewire autophagic networks remains to be elucidated.

Autophagosome elongation and maturation rely on two ubiquitin-like conjugation systems. The ATG12-ATG5-ATG16L1 complex drives membrane expansion (18), whereas microtubule-associated protein 1 light chain 3 (LC3) is conjugated to phosphatidylethanolamine, converting LC3-I to membrane-bound LC3-II, a hallmark of autophagosomes (19). Mature autophagosomes fuse with lysosomes to form autolysosomes, where engulfed cargo is degraded and recycled.

Autophagy serves pivotal roles in postembryonic development, cell differentiation and modulation of the inflammatory response, and its dysregulation is implicated in various diseases (20,21). In LC, autophagy exerts context-dependent effects. For example, ailanthone blocks autophagy by upregulating growth arrest specific (GAS)5 via inhibition of up-frameshift protein 1-mediated nonsense-mediated mRNA decay, thereby inhibiting NSCLC proliferation (22). Conversely, regulator of G protein signaling 20 facilitates NSCLC proliferation by activating autophagy through suppressing the PKA-Hippo pathway (23). Furthermore, the long non-coding (lnc)RNA KCTD21-AS1, modified by METTL14-mediated N6-methyladenosine methylation, acts as a competing endogenous RNA for miR-519d-5p to regulate CD47 and TIPRL expression, thereby modulating macrophage phagocytosis and autophagy in LC cells (24). In small cell LC (SCLC), elevated AMBRA1 promotes CDK6 degradation via autophagy and confers a favorable prognosis in patients (25). These studies collectively argue against targeting autophagy as a monotherapy; instead, they argue for a precision medicine approach based on tumor genotype and autophagic cargo specificity.

Ferroptosis is a distinct form of PCD driven by iron-dependent lipid peroxidation. Its core mechanisms involve dysregulated iron metabolism, excessive lipid peroxidation and compromised antioxidant defense (26).

Cellular iron uptake is primarily mediated by the transferrin receptor. Intracellularly, iron is stored in ferritin to maintain homeostasis; dysregulation increases the labile iron pool, which catalyzes the Fenton reaction to generate hydroxyl radicals that initiate lipid peroxidation (27). This process is central to ferroptosis: Polyunsaturated fatty acids within cellular membranes are peroxidized to form lipid hydroperoxides (LOOH). Under normal conditions, the glutathione (GSH)-GPX4 axis reduces LOOH to nontoxic alcohols, preserving membrane integrity (28). Depletion of GSH or inactivation of GPX4 leads to LOOH accumulation, resulting in the buildup of toxic peroxidation products, consequent membrane damage, and cell death (29). Other antioxidant systems, including superoxide dismutase and catalase, provide additional protection against oxidative stress. However, the clinical translatability of these in vitro findings remain equivocal owing to the absence of corroborative human tissue or cohort data.

In NSCLC, ferroptosis is regulated by diverse molecular pathways. For example, exosomal ROR1-AS1 from cancer-associated fibroblasts suppresses ferroptosis by stabilizing solute carrier family 7 member 11 (SLC7A11) mRNA via insulin-like growth factor 2 mRNA binding protein 1, thereby enhancing cystine uptake and GSH synthesis (30). Similarly, lysosome associated protein transmembrane 4B (LAPTM4B) blocks ferroptosis by suppressing NEDD4L/ZRANB1-mediated ubiquitination and degradation of SLC7A11 (31). Conversely, other mechanisms promote ferroptosis. Combined suppression of STAT3 and HOXC10 induces ferroptosis, effectively suppressing bone metastasis in KRAS-mutant LC (32). Oncogenic RIT1 mutations not only activate canonical RAS/MAPK and PI3K/AKT pathways but also modulate nuclear factor, erythroid 2 like 2 (NRF2) target expression, markedly increasing cellular susceptibility to ferroptosis inducers (33). Notably, a number of ferroptosis-regulating pathways intersect with autophagy networks, forming a complex interactive landscape.

While the core frameworks of autophagy and ferroptosis are established, their functional heterogeneity in LC presents notable dichotomies. The determinants of this duality, such as upstream signaling pathways or driver mutations, remain unclear. In ferroptosis, GPX4, typically a suppressor, paradoxically sensitizes GPX4-overexpressing NSCLC cells to RAS selective lethal 3 (RSL3)-induced ferroptosis (34). Similarly, SLC7A11 can be upregulated by LAPTM4B to inhibit ferroptosis or downregulated by circPOLA2 to promote it (35), yet the contextual switches governing its function are unknown.

These contradictions highlight critical questions: How do LC-specific adaptations of autophagy and ferroptosis differ from those in normal lung epithelium? Can these differences be therapeutically exploited? Future research should employ subtype-specific models and single-cell multi-omics to map molecular signatures, clarify how driver mutations shape pathway heterogeneity and validate these insights for targeted therapy development.

Autophagy can act as a context-dependent negative regulator of ferroptosis, a pathway frequently dysregulated in LC that fosters tumor cell survival and therapy resistance. By degrading lipid peroxidation-derived peroxisomes, misfolded proteins and damaged organelles, autophagy helps maintain redox homeostasis and suppresses ferroptosis (36,37).

This regulatory interplay involves crosstalk with other cell death pathways. Autophagy can modulate the expression and activity of apoptosis-related proteins, thereby indirectly altering cellular susceptibility to ferroptosis (38). Furthermore, mechanical tension has been shown to govern iron metabolism via nuclear receptor coactivator 4 (NCOA4)-FTH1 phase-separation-mediated autophagy, influencing ferroptosis sensitivity (39).

mTOR serves as a central nutrient and energy sensor within this network (40). Under nutrient-replete conditions, activated mTOR suppresses autophagy and strengthens antioxidant defenses, thereby restraining ferroptosis. Conversely, stress conditions inhibit mTOR, triggering autophagy and metabolic reprogramming that sensitize cells to ferroptosis. In LC, Plin2 drives NSCLC proliferation by inhibiting autophagy via AKT/mTOR activation (41), whereas circFAM190B propels tumor progression by suppressing autophagy through the SFN/mTOR/ULK1 axis (42). The compound (+)-anthrabenzoxocinone inhibits PI3K/AKT/mTOR signaling, inducing cell cycle arrest, apoptosis and autophagy, while elevating reactive oxygen species (ROS) in NSCLC cells (43). However, the specific downstream effectors of PI3K/AKT/mTOR that mediate autophagy-ferroptosis crosstalk, along with potential subtype-specific differences, require further definition.

Mitophagy, the selective autophagic clearance of mitochondria, serves a key inhibitory role against ferroptosis (44). During mitophagy, mitochondria undergo characteristic morphological changes, including swelling, cristae rupture and vacuolization, which are distinct from the cristae reduction and fragmentation observed in ferroptosis (45). Tumor cells deficient in mitophagy exhibit heightened susceptibility to ferroptosis inducers due to the loss of this protective mechanism (46). Molecules such as fibroblast growth factor 21 (47), BCL2 interacting protein 3/NIP3-like protein X (48) and acteoside (via the NRF2-mitophagy axis) (49) mitigate ferroptosis by restoring mitochondrial function or removing damaged mitochondria, underscoring the protective role of mitophagy against ferroptosis.

Under specific conditions, autophagy can promote ferroptosis, a switch often associated with malignant progression in LC. Overactivation of ferritinophagy drives rapid ferritin degradation, releasing free iron that catalyzes lipid peroxidation and induces ferroptosis (50,51). Additional promoting mechanisms include Par-4-stimulated, NCOA4-mediated ferritinophagy (52) and autophagic degradation of GSH reductase, which reduces GSH synthesis and antioxidant capacity, further promoting ferroptosis (53).

In LC, depletion of ubiquitin-specific peptidase (USP)13 facilitates the transition from autophagy to ferroptosis in KRAS-mutant LUAD cells via the NRF2-p62-kelch like ECH associated protein 1 (KEAP1) axis (54). Rapamycin, a classical autophagy inducer, can exacerbate mitochondrial damage and provoke ferroptosis when combined with multi-walled carbon nanotubes (55). Beclin-1 also induces autophagy-triggered ferroptosis by upregulating ATG5 (56).

Cytochrome c oxidase subunit 7A1 (COX7A1) exhibits a distinct dual role: It enhances mitochondrial metabolism to sensitize NSCLC cells to cysteine deprivation-induced ferroptosis, while concurrently inhibiting autophagy to block mitochondrial remodeling, two seemingly opposing effects. Rapamycin can reverse COX7A1-mediated autophagy blockade, synergistically enhancing ferroptosis under cysteine deprivation (57). The molecular switch controlling the opposing functions of COX7A1remains unknown, limiting its therapeutic exploitation. The core mechanisms of autophagy, ferroptosis and their crosstalk are summarized in Fig. 1.

Ferroptosis, through its associated oxidative stress and iron dysregulation, can reciprocally modulate autophagy, forming a dynamic bidirectional network crucial for cell fate decisions in the tumor microenvironment.

Under ferroptotic stress, JNK activation leads to phosphorylation of Beclin-1 at Ser90/93, strengthening its interaction with the VPS34 complex and stimulating autophagosome formation (58,59). This constitutes an adaptive response to ferroptosis-induced damage.

Iron overload enhances p62 phosphorylation, which strengthens its binding to KEAP1 and leads to NRF2 release and nuclear translocation. NRF2 then transcriptionally activates ATG genes, helping to mitigate ROS and maintain homeostasis (60-62). A critical unresolved question is how NRF2 balances autophagy activation and ferroptosis mitigation in the LC microenvironment, as this balance may determine tumor cell fate and therapeutic response.

Ferroptosis-associated oxidative stress also influences post-translational modifications of ATG proteins, fine-tuning their activity and localization (63-65). Ubiquitin ligase subunit FBXO9 blocks V-ATPase assembly to impede LC metastasis (66). Additionally, ferroptosis-derived ROS can activate the AMPK/ULK1 pathway to initiate autophagy (67), although the precise molecular cascade in LC cells is not fully elucidated. In summary, the crosstalk between autophagy and ferroptosis is governed by a complex, context-dependent network encompassing three core modes: i) Negative regulation, primarily through mTOR signaling and mitophagy, to inhibit ferroptosis; ii) positive regulation, driven by ferritinophagy overactivation, degradation of antioxidant proteins and regulators such as USP13, to promote ferroptosis; iii) feedback regulation, involving the JNK/Beclin-1, NRF2/p62 and ROS/AMPK pathways, creating a bidirectional adaptive loop. A defining feature is the dual role of autophagy, which can either suppress or induce ferroptosis depending on the genetic background, LC subtype and microenvironmental cues. Critical unresolved issues include identifying the molecular switches that govern the transition of autophagy from an anti- to a pro-ferroptotic state, elucidating subtype-specific differences in crosstalk mechanisms, and clarifying the molecular basis underlying proteins such as COX7A1 to exhibit opposing functions. Addressing these questions is essential for deciphering LC cell survival strategies and developing targeted therapies that exploit this dynamic interplay.

The nutrient-deprived and hypoxic tumor microenvironment of LC frequently drives the exploitation of autophagy as a cytoprotective mechanism that counteracts ferroptosis. This mechanism maintains redox homeostasis, enabling cancer cell proliferation under hostile microenvironmental stress. Genetic or pharmacological inhibition of core autophagy machinery undoes this protective mechanism, markedly sensitizing LC cells to ferroptosis inducers. For example, in cells expressing an active Rab37 mutant, knockdown of ATG5/7 not only reduces autophagic flux but also diminishes tissue inhibitor of metalloproteinase 1 secretion, leading to suppressed lung nodule formation and metastasis (68). This evidence underscores the effects of autophagy as a critical adaptive response, and demonstrates that its disruption unveils a latent vulnerability to ferroptotic death.

The regulatory interplay between autophagy and ferroptosis notably influences both the survival and metabolic processes of LC cells, particularly through metabolic remodeling. A key adaptation is the enhancement of fatty acid oxidation; a process associated with aggressive proliferation and poor patient prognosis (69,70). Dysregulated fatty acid metabolism fuels proliferation and confers targeted-therapy resistance in liver tumors; whether an analogous axis operates in LC, and whether it is further modulated by autophagy and ferroptosis, remains to be investigated (71). Furthermore, lipid metabolism enzymes directly interface with this axis. For example, stearoyl-CoA desaturase 1 suppresses autophagy by tuning lipid peroxidation and the mTOR pathway, thereby promoting NSCLC proliferation and metastasis (72). This suggests that metabolic rewiring via lipid pathways is a non-canonical mechanism by which the autophagy-ferroptosis balance is tilted toward tumor promotion.

Beyond cell-intrinsic effects, this crosstalk actively sculpts an immunosuppressive microenvironment. Autophagy can induce the secretion of immunosuppressive cytokines and regulate the expression of immune checkpoints such as PD-L1 (73). It also facilitates crosstalk with tumor-associated macrophages, particularly the M2 phenotype, to further dampen antitumor immunity. Consequently, strategic inhibition of autophagy has been shown to restore antigen presentation and synergize with immunotherapy in preclinical NSCLC models (74). These findings suggest that high baseline autophagic activity, particularly when coupled with elevated checkpoint expression, could serve as a biomarker for identifying patients who may benefit from combining autophagy inhibitors with immunotherapy, a hypothesis awaiting robust clinical validation.

Traditional pharmacological agents have demonstrated renewed potential in LC by modulating the autophagy-ferroptosis axis through diverse and sometimes unexpected mechanisms.

Antimalarial agents CQ and HCQ are repurposed for LC by blocking autophagosome-lysosome fusion, thereby blocking autophagic flux (82). Their therapeutic efficacy in LC is markedly subtype-dependent, a critical factor for clinical translation. In NSCLC, HCQ shows promise as both a chemosensitizer and immunomodulator, potentially by reducing lysosomal drug sequestration and boosting CD8⁺ T-cell responses (83).

Conversely, in a previous study on extensive-stage SCLC, the addition of HCQ to platinum-based chemotherapy failed to improve overall survival and even increased toxicity (84), underscoring that HCQ is not a universal agent and thus necessitates careful patient stratification. Nanodelivery strategies, such as encapsulation in human ferritin nanocages, can potentiate the antitumor activity of HCQ, particularly in cancer with elevated transferrin receptor 1 expression (85). Furthermore, in LKB1/KRAS co-mutant NSCLC, combining HCQ with the MEK inhibitor trametinib synergizes to induce ferroptosis. Paradoxically, autophagy induction under different conditions can confer trametinib resistance by suppressing ferroptosis (86). This bidirectional effect highlights the necessity for context-specific modulation of the autophagy-ferroptosis axis.

Classic ferroptosis inducers such as erastin can surmount radioresistance in NSCLC by triggering ferroptosis (87). The WEE1 G₂ checkpoint kinase inhibitor AZD1775 also blocks cystine uptake and synergizes with SLC7A11 inhibitors to amplify ferroptotic cell death (88). These findings validate the therapeutic targeting of the system Xc⁻ pathway, although challenges related to tumor selectivity and potential systemic toxicity remain.

Other agents function at the intersection of multiple cell death pathways. The quinoline derivative DFIQ blocks autophagic flux, leading to mitochondrial dysfunction and thereby rendering cells more susceptible to ferroptosis (89). Rabeprazole produces analogous effects and concurrently upregulates markers of pyroptosis, hinting at an uncharacterized tripartite crosstalk among autophagy, ferroptosis and pyroptosis (90). In NSCLC, carbon monoxide (CO) directly triggers ferroptosis through the ROS/GSK3β/GPX4 axis, resulting in the accumulation of LOOH (91). These findings establish the ROS/GSK3β/GPX4 axis as a key driver of CO-induced ferroptosis, but how this pathway intersects with other cell-death pathways remains to be clarified.

Reprogramming cellular metabolism presents another option to adjust ferroptosis sensitivity. Metformin, known primarily for stimulating autophagy and apoptosis via the EGFR/AKT/AMPK/mTOR pathway in SCLC (92), may indirectly influence ferroptosis given the central role of the AMPK/mTOR hub in this crosstalk, although direct evidence is required.

Targeting lipid metabolism is a promising strategy for reversing therapy resistance. The co-administration of ferroptosis inducers and diacylglycerol acyltransferase inhibitors suppresses tumor growth in LC models resistant to multiple chemotherapeutic agents (93). Similarly, modulating homocysteine metabolism through hydrogen sulfide increases ferroptosis susceptibility in NSCLC, although the precise mechanistic link requires further definition (94). These findings collectively validate 'metabolic remodeling' as a viable approach to overcome resistance. A key future direction involves defining optimal drug ratios and identifying predictive biomarkers to guide patient selection for such metabolically targeted combination therapies.

Beyond repurposing traditional drugs, the development of novel agents specifically designed to target the autophagy-ferroptosis axis represents a novel option in LC therapy. These efforts focus on distinct regulatory nodes within this interactive network.

Research into natural compounds has identified modulators of autophagic signaling as potential antitumor agents in LC. These agents converge on key regulatory nodes: The AMPK/mTOR axis and the Sestrin-2/LKB1/AMPK cascade to provoke autophagy and cell death (95-97). Others, such as corynoxine, target upstream phosphatases such as protein phosphatase 2A to dually inhibit AKT signaling (98), whereas Gboxin and its analog Y9 trigger lysosomal dysfunction and autophagic impairment (99). However, a critical gap across these studies is the general lack of investigation into whether and how these autophagy-modulating agents affect ferroptosis, which limits the assessment of their full relevance to the core crosstalk theme. Furthermore, the tumor selectivity and potential off-target toxicity of agents such as Gboxin on normal lung epithelium remain unclear, posing challenges for clinical translation.

Drug development has also progressed towards core components of the ferroptosis pathway and its key intersections with autophagy. Targeting the upstream autophagy initiator ULK1/2 presents a strategic node. The selective ULK1/2 inhibitor, DCC-3116, synergizes with the KRAS^G12C inhibitor sotorasib to suppress KRAS-driven LC, whereas this combination has been assessed only in early preclinical stages and lacks predictive biomarkers for patient stratification (17). By contrast, hesperetin inhibits ferritinophagy and mitigates ferroptosis by modulating the PI3K/AKT/mTOR/ULK1 pathway (100), further underscoring the context-dependent role of ULK1.

Inhibition of the antioxidant transcription factor NRF2 is another validated strategy to enhance ferroptosis sensitivity. The NRF2 inhibitor, ML385, synergizes with silymarin (101), and NRF2 knockdown facilitates S-3'-hydroxy-7',2',4'-trimethoxyisoflavone-mediated ferritinophagy (102), revealing a reciprocal regulatory loop between NRF2 and autophagy. Other agents include sanggenol L, which acts through the miR-26a-1-3p/MDM2/p53/SLC7A11 axis (103), and curcumin, which activates autophagy to subsequently trigger ferroptosis (75). Polyoxometalates exhibit synergistic anti-NSCLC activity with ferroptosis and apoptosis inducers (104), although the specific nodes of crosstalk involved remain to be defined.

To address the pharmacokinetic and selectivity limitations of classic ferroptosis inducers, advanced formulation strategies are being explored. A folic acid-modified liposomal nanoformulation co-delivering erastin and the metallothionein 1D pseudogene markedly augments bioavailability and cytotoxicity in LC (105). Concurrently, understanding resistance mechanisms is crucial. Upregulation of the SLC7A11/HOXB9 axis can mitigate cold atmospheric plasma-induced ferroptosis while fueling tumor progression (106). Conversely, inducing ferroptosis sensitizes LC cells to IFN-γ secreted by ROR1-targeted chimeric antigen receptor (CAR)-T cells, revealing a potent synergy with immunotherapy (107). A recurring limitation in both therapeutic development and resistance studies is the failure to consider the potential contributory role of autophagy, leaving the integrated network perspective incomplete.

Notable challenges persist in translating the autophagy-ferroptosis axis into clinical applications for LC, despite growing mechanistic understanding. The primary hurdle is the intrinsic complexity and context-dependency of this interplay. Its regulatory network remains incompletely mapped, as it involves numerous overlapping and often contradictory signaling pathways. A critical gap is the identification of definitive molecular switches that determine whether the crosstalk promotes tumor cell survival or death in a given therapeutic setting, which severely impedes the rational design of pathway-specific modulators.

Substantial inter- and intra-tumoral heterogeneity, epitomized by the profound biological divergence between LUAD and SCLC (149), remains a major obstacle to developing universal treatment paradigms. Although multimodal regimens combining radiotherapy, chemotherapy and immunotherapy have achieved notable clinical benefits across LC subtypes (150), the molecular drivers of these differential responses remain poorly understood. This knowledge gap represents a critical barrier to advancing truly precision-based therapeutic strategies.

Most therapeutic interventions remain in preclinical development, with clinical translation hindered by two key barriers: The lack of validated biomarkers for patient stratification and incomplete elucidation of synergistic combination mechanisms. A central, unresolved question remains how to precisely balance the modulation of autophagy and ferroptosis to maximize therapeutic efficacy while minimizing the risks of tumor promotion or off-target tissue damage.

To tackle these barriers, future research should prioritize the following key areas.

Employing integrated multi-omics approaches across diverse LC subtypes and dynamic treatment states is essential. This will enable the construction of a high-resolution map of the molecular landscape, the identification of subtype-specific regulatory nodes and switches (151), and address the fundamental knowledge gaps highlighted in preceding sections, thereby laying a foundation for precise therapeutic targeting.

Predictive biomarkers, particularly those linked to key regulatory axes such as NRF2/p62, COX7A1 or USP13, require urgent discovery and clinical validation to reliably predict sensitivity to ferroptosis inducers or autophagy modulators. Correlating longitudinal multi-omics data from large, well-annotated patient cohorts with clinical outcomes will thus enable accurate patient stratification, guide therapeutic decision-making, and increase the success rate of clinical trials.

Advanced computational approaches, including AI and machine learning algorithms, are increasingly employed to decipher the heterogeneity of autophagy-ferroptosis regulation and optimize therapeutic decisions by integrating multimodal data (genomics, transcriptomics, proteomics and clinical data) (152,153). Their core value is concentrated in three key directions, as follows: i) Deep learning, network-based machine learning and other algorithms are utilized to mine patient-specific autophagy-ferroptosis molecular signatures, enabling precise stratification of treatment responders and non-responders, and providing a basis for individualized therapeutic target selection (153). ii) AI models trained on longitudinal clinical and molecular data integrate molecular features and clinical variables to predict individual sensitivity to specific therapeutic regimens, reducing reliance on therapeutic trial and error, and optimizing drug dosage and sequence (154). iii) AI-driven real-time analytical technologies monitor dynamic reprogramming in crosstalk pathways during treatment, guiding timely adjustments to therapeutic strategies and effectively overcoming inter-patient variability in treatment responses among advanced patients (155,156).

These technologies may accelerate the discovery of novel therapeutic targets and the development of personalized regimens, serving as critical cutting-edge tools to advance autophagy-ferroptosis-targeted therapy from basic research to clinical translation.

Building upon the mechanistic foundations, translational efforts must now focus on: i) Optimizing the pharmacokinetics, biodistribution and safety profiles of advanced delivery systems; ii) conducting large-scale, biomarker-driven clinical trials to evaluate rational combination therapies; and iii) exploring the underexplored role of the autophagy-ferroptosis crosstalk in critical clinical processes such as metastasis, tumor dormancy and therapy-resistant recurrence.