Autophagy is orchestrated by a series of autophagy-related genes (ATGs) that encode protein complexes acting in concert. The autophagy process can be divided into several key stages: (i) induction or initiation; (ii) cargo selection; (iii) phagophore nucleation; (iv) phagophore membrane elongation and expansion; (v) retrieval and fusion with lysosomes; and (vi) degradation of intravesicular contents. Blocking autophagy at different stages may result in distinct biological effects and outcomes [1]. Adenosine triphosphate (ATP)/adenosine monophosphate (AMP) imbalance is one important trigger of autophagy [2]. The Unc-51-like kinase 1 (ULK1) complex is essential for autophagy initiation. ULK1 activity is finely regulated by upstream signals such as AMP-activated protein kinase (AMPK) [3] and by inhibitory signaling through the mechanistic target of rapamycin complex 1 (mTORC1) [4]. Downstream of the ULK1 complex, the coiled-coil myosin-like BCL2-interacting protein (Beclin 1) complex generates phosphatidylinositol-3-phosphate (PI3P), which facilitates the recruitment of other ATG proteins [5]. The nucleation stage of autophagosome formation involves interactions among Beclin 1-PI3KC3 (class III PI3K complex), Rubicon, Ambra1, and other factors [6]. Subsequent membrane elongation and closure depend on two ubiquitin-like conjugation systems: the ATG12–ATG5 conjugation system and the ATG8 conjugation system. The latter mediates the lipidation of ATG8 family members, including microtubule-associated protein 1 light chain 3 (LC3) [7]. LC3 serves as a structural component on the membranes of phagophores and autophagosomes [8], and it exists in three isoforms (LC3A, LC3B, and LC3C). Because LC3 is typically incorporated into both the inner and outer membranes of autophagosomes [9], it is widely used as a marker for monitoring autophagy [10].

In NSCLC, dysregulation of mTORC1—a master nutrient and growth regulator—features prominently and intersects with ULK1-driven initiation to shape autophagic flux in a mutation- and context-dependent manner (Fig. 1). Studies have shown that inhibitor of apoptosis-stimulating protein of p53 (iASPP) is overexpressed in human NSCLC, and knockout of iASPP can block autophagosome formation via inhibition of the mTORC1-p70 ribosomal S6 kinase (p70S6K) signaling pathway, thereby impairing autophagy. Conversely, iASPP overexpression induces autophagic flux and promotes NSCLC cell growth and tumorigenesis. Circular RNAs (circRNAs) with 5-methylcytosine (m5C) modifications have been reported to suppress autophagy through the stratifin (SFN)/mTOR/ULK1 pathway, ultimately promoting lung cancer progression [11]. A newly identified regulatory pathway, the argonaute-4 (AGO4)–tripartite motif-containing protein 21 (TRIM21)–78-kDa glucose-regulated protein (GRP78) axis, induces apoptosis and inhibits autophagy by activating the mTOR signaling pathway, representing a potential therapeutic target for treating p53-deficient tumors [12]. Ubiquitin-conjugating enzyme E2T (UBE2T) has been shown to upregulate autophagy in NSCLC cells via activation of the p53/AMPK/mTOR signaling pathway [13]. Fas apoptosis inhibitory molecule 1 (FAIM1) can promote the binding of ULK1 to mTOR, which in turn suppresses autophagy induction [14].

Fig. 1.

Integrated signaling and metabolic regulation of autophagy in lung cancer. The figure illustrates key molecular pathways that regulate autophagy in lung cancer cells, emphasizing the central role of mTORC1 as an integration hub for nutrient, growth factor, and metabolic signals. 4E-BP1, Eukaryotic translation initiation factor 4E-binding protein 1; ADSL, Adenylosuccinate lyase; AKT, Protein kinase B; AMBRA1, Activating molecule in Beclin-1-regulated autophagy 1; AMP, Adenosine monophosphate; AMPK, AMP-activated protein kinase; ATP, Adenosine triphosphate; ATG3, Autophagy related 3; ATG5, Autophagy related 5; ATG12, Autophagy related 12; ATG16L1, Autophagy related 16 like 1; Atg13, Autophagy related 13; circRAPGEF5, Circular RNA derived from RAPGEF5; DCP2, mRNA-decapping enzyme 2; FIP200, FAK family-interacting protein of 200 kDa (RB1CC1); GLUTs, Glucose transporters (GLUT/SLC2A family); GTP, Guanosine triphosphate; HACE1, HECT and ankyrin repeat containing E3 ubiquitin protein ligase 1; HMBOX1, Homeobox containing 1; IRS1, Insulin receptor substrate 1; LKB1, Liver kinase B1 (STK11); LC3B-I, Microtubule-associated protein 1 light chain 3 beta, non-lipidated form; LC3B-II, Microtubule-associated protein 1 light chain 3 beta, lipidated form; m6A, N6-methyladenosine; METTL3, Methyltransferase-like 3; mTOR, Mechanistic target of rapamycin; mTORC1, Mechanistic target of rapamycin complex 1; NRBF2, Nuclear receptor binding factor 2; p53, Tumor protein p53; p62, Sequestosome-1 (SQSTM1); p70S6K, 70-kDa ribosomal protein S6 kinase; p300, Histone acetyltransferase p300 (EP300); PDK1, 3-Phosphoinositide-dependent protein kinase-1; PI3K, Phosphoinositide 3-kinase; PI3KC3, Class III phosphatidylinositol 3-kinase complex (PIK3C3/Vps34 complex); PI103, PI-103 (PI3K/mTOR inhibitor); PIP2, Phosphatidylinositol (4,5)-bisphosphate; PIP3, Phosphatidylinositol (3,4,5)-trisphosphate; Rag, Rag GTPases (RRAGA/B and RRAGC/D); Rheb, Ras homolog enriched in brain (small GTPase); SCD1, Stearoyl-CoA desaturase 1; SLC1A5, Solute carrier family 1 member 5 (ASCT2); SLC3A2, Solute carrier family 3 member 2; SREBF1, Sterol regulatory element-binding transcription factor 1 (gene encoding SREBP1); SREBP1, Sterol regulatory element-binding protein 1; TCA, Tricarboxylic acid cycle; TFEB, Transcription factor EB; TSC1/2, Tuberous sclerosis complex 1/2; ULK1, Unc-51-like autophagy activating kinase 1; Vps15, Phosphoinositide 3-kinase regulatory subunit type 4 (PIK3R4); Vps34, Class III phosphatidylinositol 3-kinase (PIK3C3); WIPI2, WD repeat domain phosphoinositide-interacting protein 2.The dashed arrow represents transmembrane regulation, and the solid arrow represents direct regulation.

Alterations in Beclin 1, a core autophagy gene, are associated with lung cancer prognosis. Low expression of Beclin 1 has been linked to poor outcomes through diverse mechanisms. Ubiquitin-specific proteases (USPs) can promote NSCLC progression by deubiquitinating and stabilizing oncogenic target proteins [15]. For example, kirsten rat sarcoma viral oncogene homolog (KRAS) signaling increases reactive oxygen species (ROS) production, leading to the dimerization and stabilization of USP5. Activated USP5 deubiquitinates and stabilizes Beclin 1, which enhances autophagy and concurrently promotes the degradation of tumor suppressor p53. Inhibition of USP5 or Beclin 1 suppresses autophagy and effectively blocks KRAS-driven NSCLC growth [16]. Similarly, disruption of Toll-like receptor 4 (TLR4)-induced TRAF6–Beclin 1 signaling can suppress autophagy, resulting in reduced cancer cell migration and invasion [17]. USP15 is significantly downregulated in primary NSCLC (especially lung adenocarcinoma), and its expression is inversely correlated with disease progression. Mechanistically, USP15 positively regulates autophagy induction through the TRAF6–Beclin 1 signaling axis, thereby acting as a negative regulator of lung cancer progression [18]. Aberrant RNA splicing has also been implicated in autophagy regulation in lung cancer. Splicing factor serine/arginine-rich splicing factor 6 (SRSF6) is frequently amplified in lung cancer and modulates the splicing of transcripts involved in cancer progression [19]. Likewise, depletion of another splicing factor, SRSF1, has been found to activate autophagy. Mechanistically, SRSF1 promotes the splicing of Bcl-x to produce the long isoform Bcl-xL, which can bind Beclin 1 and inhibit autophagy; reduction of SRSF1 leads to increased production of the short isoform Bcl-xS, relieving this inhibition and thereby inducing autophagy. Tripartite motif-containing protein 59 (TRIM59) has been identified as a potential oncogene in NSCLC [20]. TRIM59 expression is negatively correlated with Beclin 1 levels, and TRIM59 knockdown significantly elevates basal autophagy in NSCLC cells. TRIM59 appears to regulate autophagy by downregulating the NF-$\kappa$B pathway, thereby reducing BECN1 (Beclin 1 gene) transcription. An E3 ubiquitin ligase known as glycogenin-interacting protein 1 (GXRNIP1 also referred to as GNIP1) is overexpressed in lung tumor tissues, and higher GNIP1 levels are associated with poor prognosis in NSCLC patients. In NSCLC cells, GNIP1 facilitates the recruitment of Beclin 1 and LC3B to form autophagosomes, promoting cancer cell proliferation and migration [21, 22, 23].

SOX2, a transcription factor, is highly expressed in lung cancer cells and exhibits oncogenic properties, contributing to tumorigenesis and therapy resistance [24]. There is evidence of crosstalk between autophagy and SOX2 signaling. Specifically, downregulation of LC3A (one of the LC3 isoforms) can inhibit lung cancer cell growth; the interaction between LC3A-mediated autophagy and SOX2 proliferative signaling suggests that autophagy may enhance SOX2-driven proliferation in lung cancer cells [25].

Autophagy exerts both tumor-suppressive and tumor-promoting effects in lung cancer, depending on the stage of cancer development. In the early stages of lung tumorigenesis, autophagy predominantly acts as a tumor suppressor through several mechanisms: maintaining genomic stability by removing damaged organelles and proteins, promoting programmed cell death (including autophagy-dependent cell death and apoptosis) in aberrant cells, and participating in immune surveillance to eliminate nascent tumor cells. Thus, during early cancer development, inhibition of autophagy may unintentionally accelerate tumor growth. By contrast, at later stages of cancer, tumor cells often hijack autophagy to support survival, thereby overriding its tumor-suppressive effects. This stage-dependent switch means that blocking core autophagy regulators could have opposite effects at different times: potentially promoting tumorigenesis if applied too early, but impairing tumor cell viability in established cancers.

Early in tumorigenesis, autophagy mainly preserves genome and organelle integrity; once malignancy is established, the same machinery becomes a survival toolkit under hypoxia and therapy stress. The complex effect of autophagy on tumor dynamics is exemplified by studies on ATG7, a key autophagy gene. Deletion of Atg7 in mouse models initially accelerated lung tumor development, but at later stages it inhibited further tumor progression [26]. One proposed mechanism for this phenomenon is that Atg7-deficient tumor cells are unable to efficiently remove dysfunctional mitochondria, leading to excessive ROS accumulation. Elevated ROS can promote tumor initiation by increasing DNA damage and mutations, but persistent oxidative stress and damage in established tumors may eventually limit growth or viability of cancer cells [27]. The tumor suppressor p53 has also been shown to regulate autophagy in multiple ways. Notably, p53’s effect on autophagy depends on its subcellular localization. Nuclear p53 can activate autophagy by transcriptionally upregulating autophagy-related genes (such as the Sestrin family), and cellular stress that stabilizes p53 often leads to the activation of genes that promote autophagy [28]. In contrast, cytoplasmic p53 can inhibit autophagy, in part through its effects on AMPK and mTOR signaling [29]. In lung cancer stem cells, autophagy has been reported to enhance stemness by degrading ubiquitinated p53, thereby alleviating the autophagy-inhibitory effect of cytosolic p53 [30]. Sestrin2, a conserved oxidative stress sensor and metabolic regulator downstream of p53, can strongly suppress oncogenic pathways by downregulating mTORC1 and hypoxia-inducible factor 1$\alpha$ (HIF-1$\alpha$), linking p53 activity to metabolic control of tumor growth [31, 32].

Autophagy also plays a role in anti-tumor immune responses. It can promote the presentation of tumor antigens and thus enhance both innate and adaptive immune surveillance of cancer cells. Immune checkpoint inhibitors have become a first-line immunotherapy for programmed cell death ligand 1 (PD-L1)-positive non-squamous NSCLC patients, highlighting the importance of immune evasion mechanisms in lung cancer therapy [33, 34]. Intriguingly, autophagy has been found to regulate the expression and turnover of immune checkpoints such as PD-1 and its ligand PD-L1 [35]. The selective autophagy adaptor protein p62/SQSTM1 serves as an immunomodulator; p62-mediated selective autophagy can potentiate innate immune responses by modulating signaling pathways. In NSCLC, p62 accumulation has been shown to regulate PD-L1 levels by binding directly to PD-L1 and targeting it for lysosomal degradation, thereby potentially enhancing the anti-tumor immune response [36]. Natural killer (NK) cells are critical effectors in tumor immunosurveillance [37]. It has been demonstrated that chemokines like CXCL10 play a vital role in directing NK cell homing and infiltration into solid tumors and in enhancing their cytotoxic efficacy [38]. This suggests that inducing tumor cells to secrete chemoattractants such as CXCL10 and CCL5 via autophagy-related pathways could be an effective strategy to recruit NK cells and boost anti-tumor immunity.

As lung cancer progresses to advanced stages, the functional role of autophagy often shifts from tumor suppression to tumor promotion. Within the nutrient- and oxygen-deprived tumor microenvironment, autophagy becomes a survival mechanism, providing energy and metabolic substrates through the lysosomal degradation of cellular components. Loss-of-function mutations in tumor suppressors that normally restrain autophagy can thus confer a growth advantage to tumors. For instance, inactivating mutations of TP53 are among the most common genetic alterations in NSCLC [39]. TP53 loss leads to the indirect activation of multiple compensatory pathways that support tumor progression, many of which involve upregulation of autophagy. One such pathway is the liver kinase B1-AMP-activated protein kinase (LKB1–AMPK) signaling cascade, which, when activated, promotes tumor growth by enhancing autophagy-dependent metabolism [40]. Notably, approximately 50% of KRAS-mutant NSCLCs occur in the context of concurrent LKB1 loss-of-function mutations (the so-called “KL” genotype) [41]. LKB1 loss has been shown to induce the transcription of NEDD9, a scaffolding protein implicated in metastasis and autophagy regulation [42]. Recent findings indicate that in NSCLC cells deficient in LKB1 and NEDD9, there is a post-transcriptional induction of autophagy through the LKB1–AMPK signaling axis, which supports the survival of these aggressive cancer cells under metabolic stress [43].

The ubiquitin-conjugating enzyme E2C (UBE2C) is another factor linking autophagy to lung cancer progression. High UBE2C expression is associated with poor overall survival in lung cancer patients [44]. In cellular models, UBE2C overexpression leads to suppression of autophagy and is correlated with enhanced cell proliferation, survival, and malignant phenotype in lung cancer cells [45]. Mechanistically, UBE2C has been shown to target the mTOR pathway regulator DEP domain containing 6 (DEPTOR) for ubiquitination and degradation, in cooperation with the E-cadherin (CDH1) complex [46, 47]. UBE2C forms a complex with CDH1 to promote DEPTOR ubiquitination; accordingly, knockdown of UBE2C causes accumulation of DEPTOR (stabilizing mTOR inhibition) and growth arrest, which can be largely reversed by co-knockdown of DEPTOR. Nuclear factor erythroid 2–related factor 2 (NRF2), a key transcription factor that upregulates a broad spectrum of antioxidant genes, is essential for KRAS-driven lung tumorigenesis, in part by mitigating oxidative stress. However, NRF2 hyperactivation can also lead to increased autophagy and has been linked to both autophagic and apoptotic cell death in NSCLC models [48].

Autophagy contributes to therapeutic resistance in lung cancer through multiple mechanisms. Hypoxic regions of the tumor stabilize HIF-1$\alpha$, which in turn activates gene expression programs that drive cancer cell proliferation, migration, chemoresistance, and immune evasion [49]. For example, the long non-coding RNA plasmacytoma variant translocation 1 (PVT1) is highly expressed in lung tumors and cell lines, particularly under hypoxic conditions. PVT1 has been implicated in hypoxia-induced chemoresistance; HIF-1$\alpha$ directly upregulates PVT1, which then promotes autophagy via the PVT1/miR-551b/FGFR1 axis, leading to increased cell viability and reduced apoptosis in the presence of chemotherapy [50]. Another study found that HIF-1$\alpha$ interacts with eukaryotic initiation factor 5A2 (eIF5A2), resulting in its overexpression during hypoxia-induced autophagy and contributing to cisplatin resistance. Silencing eIF5A2 by RNA interference attenuated hypoxia-induced autophagy and resensitized NSCLC cells to cisplatin under hypoxic conditions [51]. Additionally, HIF-1$\alpha$ can cooperate with histone deacetylase 4 (HDAC4) to dysregulate the balance between apoptosis and autophagy through intersecting p53 and RAS signaling pathways, thereby actively driving cisplatin resistance [52].

A deep understanding of the stage-specific roles of autophagy in lung cancer is crucial for developing precise autophagy-targeted therapeutic strategies. In the following sections, we will explore the intricate interplay between autophagy and metabolic reprogramming in lung cancer, an area that has seen a series of groundbreaking advances in recent years.

Lung cancer cells exhibit profound metabolic reprogramming to meet the energy and biosynthetic demands of unchecked growth. Under conditions of nutrient starvation or hypoxia, glycolysis is often upregulated (the “Warburg effect”), and autophagy has been shown to support tumor cell survival and proliferation by sustaining aerobic glycolysis and fueling the tricarboxylic acid (TCA) cycle with recycled substrates [54]. For example, studies have demonstrated that autophagy promotes tumorigenesis in pancreatic cancer cells by maintaining glycolytic flux and cellular energy levels, while also preventing apoptosis through ROS-mediated activation of AMPK [55]. Conversely, in the tumor microenvironment, an abundance of glucose can inhibit autophagy: high glucose levels were found to upregulate SREBF1/SREBP1 (sterol regulatory element-binding protein 1), which in turn impairs autophagy and attenuates apoptosis in cancer cells. Furthermore, oxidative stress in tumor cells (excess ROS production) can induce metabolic reprogramming in cancer-associated fibroblasts (CAFs), increasing autophagy in the stromal cells and leading to the production of “fuel” molecules such as lactate, ketone bodies, fatty acids, and glutamine. These metabolites are released by CAFs and taken up by cancer cells to support tumor growth and metastasis under metabolic stress conditions [56].

To sustain the high metabolic demands, cancer cells utilize diverse nutrient sources to fuel the mitochondrial TCA cycle. Lipids represent one major alternative fuel. Through fatty acid $\beta$-oxidation (FAO), lipids can generate significant amounts of ATP, and they also provide building blocks for membrane synthesis and serve as signaling molecules [57]. Altered lipid metabolism is a hallmark of NSCLC, and autophagy plays a central role in lipid homeostasis. Autophagy regulates lipid stores through a selective form of autophagy known as lipophagy, where lipid droplets are broken down to release free fatty acids that can be utilized for energy production or membrane remodeling [58]. In parallel, tumor cells often exhibit increased de novo lipogenesis and accumulation of lipid droplets. Notably, recent research highlights the involvement of small nucleolar RNAs (snoRNAs) in linking lipid metabolism with autophagy regulation in lung cancer. A panel of plasma snoRNAs has been proposed as potential NSCLC diagnostic biomarkers [59]. In particular, SNORD88C was identified as a snoRNA that drives lipid metabolic reprogramming and autophagy inhibition in NSCLC. SNORD88C, through 2^′-O-methylation of 28S rRNA at a specific site, enhances the translation of stearoyl-CoA desaturase 1 (SCD1), an enzyme that synthesizes monounsaturated fatty acids (MUFA) [60]. As a result, elevated SCD1 levels suppress autophagy and promote lung tumor growth and metastasis. Mechanistically, SNORD88C forms a small nucleolar ribonucleoprotein (snoRNP) complex with the fibrillarin protein, which mediates 2^′-O-methylation at nucleotide C3680 of 28S rRNA, thereby promoting ribosome biogenesis and preferential translation of SCD1 mRNA [61]. SCD1, the rate-limiting enzyme for MUFA synthesis, inhibits autophagy via two distinct mechanisms: (1) the MUFAs (e.g., oleic acid) produced by SCD1 reduce lipid peroxidation levels, which indirectly dampens autophagy; and (2) SCD1 activity leads to activation of the mTOR/ULK1 pathway, directly blocking autophagy initiation [62]. This discovery reveals a novel snoRNA-driven epitranscriptomic mechanism connecting lipid metabolism and autophagy, and it suggests SNORD88C as a potential plasma biomarker for NSCLC. Conversely, autophagy can feed back into lipid metabolism by degrading lipid droplets (lipophagy), thereby releasing fatty acids for $\beta$-oxidation or for reuse in membrane synthesis.

Another recent study highlighted the tumor-suppressive role of PAQR3 (progestin and AdipoQ receptor family member 3) in NSCLC and its connection to autophagy. PAQR3 is frequently downregulated in NSCLC cells, and restoration of PAQR3 expression significantly inhibits tumor cell proliferation [63]. Mechanistically, PAQR3 re-expression was found to markedly enhance erlotinib-induced autophagy, thereby suppressing tumor growth [64]. Strikingly, when autophagy was pharmacologically inhibited, the anti-proliferative effect of PAQR3 was completely abrogated, indicating that autophagy is a critical mediator of PAQR3’s tumor suppressive function. This finding provides a fresh perspective on the anti-tumor mechanism of EGFR tyrosine kinase inhibitors (TKIs) like erlotinib, suggesting that they may exert some of their effects by perturbing tumor lipid homeostasis via a PAQR3-dependent autophagy pathway [65].

Autophagy also interfaces with ferroptosis, an iron-dependent form of programmed cell death characterized by the accumulation of lethal lipid peroxides. When lipid peroxides reach excessive levels, ferroptosis can be triggered; tumor cells often evade this form of death by suppressing processes like ferritinophagy (autophagic degradation of ferritin) that would otherwise increase free iron and promote lipid peroxidation. Indeed, evidence indicates that the autophagy machinery is critically involved in promoting ferroptosis. For instance, autophagy (particularly ferritinophagy) can facilitate erastin-induced ferroptosis by degrading ferritin and elevating intracellular iron levels, thereby enhancing lipid peroxidation [66]. Recent studies have identified factors that link mitochondrial metabolism, autophagy, and ferroptosis sensitivity in lung cancer. COX7A1, a subunit of cytochrome c oxidase in the mitochondrial electron transport chain, was found to increase NSCLC cell sensitivity to cystine deprivation–induced ferroptosis by enhancing the TCA cycle and complex IV activity, thereby altering mitochondrial metabolism and ROS production. Moreover, NSCLC cells rely on Atg7 to sustain mitochondrial fatty acid oxidation and maintain lipid homeostasis; loss of Atg7 leads to disrupted lipid metabolism and increased vulnerability to metabolic stress.

In summary, cancer cells adapt to metabolic stress not only by reprogramming metabolic pathways to support biosynthesis and maintain redox balance, but also by upregulating nutrient scavenging pathways, with autophagy being a prime example. Autophagy’s degradative function captures and degrades intracellular macromolecules (proteins, lipid droplets, organelles) and recycles them into metabolic substrates, which becomes vital when extracellular nutrients are limited [67]. For example, NSCLC cells harboring co-mutations in KRAS and LKB1 (KL subtype) exhibit an epithelial-to-mesenchymal transition (EMT)-like phenotype and notably high autophagy-lysosomal activity [68]. In these KL mutant cancers, enhanced autophagy provides intracellular citrate and acetyl-CoA, which fuel epigenetic changes; specifically, autophagy-derived acetyl-CoA was shown to promote CBP-mediated acetylation of the EMT transcription factor Snail, thereby facilitating cancer cell invasion and metastasis [69]. Selective autophagy of mitochondria (mitophagy) is particularly important for mitochondrial quality control and metabolic adaptation [70]. In small cell lung cancer (SCLC), elevated autophagy (especially mitophagy) has been linked to chemotherapy resistance; SCLC cell lines that are chemoresistant show significantly higher autophagic activity than their chemosensitive counterparts [71]. One study found that METTL3 (an m6A RNA methyltransferase) is upregulated in chemoresistant SCLC cells and promotes mitophagy, contributing to chemotherapy resistance. Inhibition of METTL3 or mitophagy partially reversed this resistance, indicating a potential therapeutic approach [72]. Indeed, mitophagy has been directly implicated in treatment resistance: excessive removal of damaged mitochondria can allow cancer cells to evade apoptosis induced by therapies. For example, brain-expressed X-linked 2 (BEX2) protein has been shown to protect cells from apoptosis by enhancing autophagic flux [73]. Likewise, S100A4 (fibroblast-specific protein 1) has been reported to promote lung cancer cell proliferation by activating Wnt/$\beta$-catenin signaling and concurrently inhibiting starvation-induced autophagy, thereby tipping the balance towards survival and growth [74]. Autophagy-deficient tumor cells often experience endoplasmic reticulum (ER) stress due to the accumulation of misfolded proteins and damaged organelles that would normally be degraded. TRAF3IP3 is one example of a protein that connects these processes: it can trigger ER stress via the PERK/ATF4/CHOP pathway, which in lung adenocarcinoma cells leads to the induction of a protective autophagy response (ER stress-induced autophagy) to cope with the unfolded protein burden [75]. In addition, a c-Myc/miR-150/EPG5 axis has been described in NSCLC whereby autophagy deficiency (caused by miR-150-mediated inhibition of the autophagy tethering factor EPG5) induces chronic ER stress, elevates ROS levels, and activates the DNA damage response, ultimately promoting cancer cell proliferation and tumor growth [76, 77]. Notably, tumor–stromal interactions can influence autophagy as well. It has been shown that lung cancer cells can co-opt normal fibroblasts into cancer-associated fibroblasts (CAFs) through factors delivered by extracellular vesicles (EVs); one study identified EV-packaged miR-1290 as a key player that transforms normal fibroblasts into CAFs [78, 79]. Moreover, these CAFs can transfer mitochondrial DNA to cancer cells via tunneling nanotubes, a process that involves autophagy-related mechanisms (termed “trogocytosis” or organelle transfer). By acquiring intact mitochondrial DNA from CAFs, cancer cells can restore mitochondrial function, better resist oxidative stress, and gain enhanced metastatic capacity. This underlines the multifaceted role of autophagy not only within cancer cells but also in modulating the tumor microenvironment to favor cancer progression.

Lipid metabolic reprogramming is a prominent feature of cancer cells and has profound effects on autophagy, thereby influencing malignant progression in NSCLC. This interaction operates at multiple levels:

Autophagy regulates lipid stores through lipophagy, wherein lipid droplets are degraded to release free fatty acids that feed $\beta$-oxidation and membrane biogenesis [80]. Lung tumors frequently exhibit increased de novo lipogenesis and lipid droplet accumulation, creating a dynamic balance between lipid storage and autophagic clearance that shapes membrane composition and signaling [81]. A notable epitranscriptomic axis links rRNA modification to lipogenesis and autophagy control: SNORD88C forms a snoRNP with fibrillarin to 2^′-O-methylate 28S rRNA, selectively enhancing SCD1 translation [61]; the resulting MUFA output reduces lipid peroxidation and activates the mTOR/ULK1 brake, thereby suppressing autophagy and promoting metastasis. This finding suggests that epitranscriptomic control can steer membrane composition and autophagic set-points in NSCLC and nominates SNORD88C–SCD1 as a tractable node. Membrane sterol homeostasis also couples to autophagy: p-toluenesulfonamide (PTS) decreases membrane cholesterol, elevates LC3-II and reduces p62, and this autophagy induction is reversed by exogenous cholesterol—implicating an actionable cholesterol–autophagy axis. In parallel, PAQR3 restoration potentiates erlotinib-induced autophagy and growth suppression, whereas pharmacologic autophagy inhibition abolishes PAQR3’s benefit, highlighting an autophagy-dependent tumor-suppressive route [62]. As a master nutrient sensor, mTOR integrates growth and lipid cues to modulate autophagy and cooperates with PPAR$\gamma$ and transcription factor EB (TFEB) to coordinate lysosome–lipid gene programs [82]. Therapeutically, co-modulating cholesterol handling (e.g., ACAT1/HMGCR) or desaturation (SCD1) alongside autophagy may rebalance membrane fluidity/vesicular trafficking and sensitize tumors to EGFR-TKIs or chemotherapy.

Mechanistic plausibility in NSCLC stems from the high purine demand and rewired central carbon metabolism of KRAS-driven—particularly KL (KRAS with LKB1 loss)—tumors, where increased ADSL flux could elevate intratumoral fumarate [83]. While this provides a biologically coherent premise, we deliberately frame ADSL–fumarate–Beclin 1 as a falsifiable hypothesis until lung-specific data demonstrate (i) coordinated increases in ADSL activity, fumarate/2-succinocysteine (2-SC) burden, and Beclin 1 dimethylation, and (ii) autophagy-flux enhancement that is not fully explained by parallel stress pathways [84].

To enable falsification rather than assumption, we outline an integrated experimental roadmap: (1) multi-modal readouts in NSCLC models and specimens (targeted metabolomics for fumarate; 2-SC immunodetection; Beclin 1 dimethylation by IP-MS or site-directed antibodies; autophagy flux via LC3-II/p62 dynamics under short lysosomal blockade and mCherry-GFP-LC3 reporting); (2) genetic and pharmacologic perturbations (CRISPR ADSL knockout/overexpression; catalytic-site mutants; rescue with cell-permeant fumarate versus non-equivalent dicarboxylate controls; Beclin 1 methylation-deficient mutants to test pathway specificity); (3) isotope tracing with [U-¹³C]-aspartate/glutamine to confirm fumarate routing into the TCA cycle and its temporal coupling to flux changes; and (4) validation in KL-biased PDX/GEMM and prospective clinical cohorts with paired on-treatment biopsies. Because KEAP1/NRF2 alterations and hypoxia can confound fumarate-responsive stress programs, analyses should be stratified by KEAP1/NRF2 status and hypoxia scores to isolate pathway-specific effects [69, 85].

Beyond ADSL, purine–autophagy crosstalk in lung cancer converges on energy and epigenetic routes already supported by our review: XDH-driven nucleotide catabolism sustains amino-acid supply and activates unfolded protein response (UPR)-linked autophagy under starvation to support LUAD survival [86]; Fat mass and obesity-associated protein (FTO)-mediated m6A demethylation upregulates PELI3, enhancing autophagy and gefitinib resistance [87]; NNT tunes autophagy through the NAD+/SIRT1 axis, shaping cisplatin sensitivity [88]; and Hsp70 can suppress AMPK and autophagy, whereas Hsp70 inhibition plus chemotherapy elicits robust (pro-death) autophagy and augments cisplatin cytotoxicity [89]. In parallel, CDK4/6 inhibitors induce autophagy and drive AMBRA1-dependent lysosomal degradation of CDK6, underscoring actionable cell-cycle–autophagy coupling and motivating dual-pathway designs that integrate purine-autophagy nodes with cycle control [90].

In sum, these observations should be regarded as graded-evidence mechanistic inferences. Only after lung-specific validation—demonstrating causal links among ADSL flux, fumarate/2-SC accumulation, Beclin 1 dimethylation, and autophagy-dependent phenotypes—should therapeutic generalization be entertained. Until then, we advocate biomarker-embedded, genotype-stratified studies to determine whether the ADSL–Beclin 1 axis represents a true autophagy dependency in NSCLC or a context-limited phenomenon.

By governing ferritinophagy (NCOA4-mediated ferritin turnover) and lipid peroxidation thresholds, autophagy can either gate or accelerate ferroptosis. Autophagic degradation of ferritin increases labile iron and amplifies lipid ROS, converting sublethal stress into ferroptotic death under defined metabolic states [91]. COX7A1-driven reinforcement of the TCA cycle and complex IV activity sensitizes NSCLC to cystine-deprivation ferroptosis, while Atg7 is required to sustain mitochondrial fatty-acid oxidation and lipid homeostasis; Atg7 loss disrupts lipid handling and increases ferroptosis vulnerability [92]. In addition to ferritinophagy, cargo selectivity determines directionality: (i) lipophagy mobilizes fatty acids and can enrich PUFA substrates in membranes via ACSL4/LPCAT3, thereby heightening peroxidation propensity, whereas SCD1-derived MUFAs dilute PUFA sites and dampen ferroptosis; (ii) mitophagy removes ROS-generating mitochondria and thus typically buffers lipid peroxidation [93]. Consequently, enhancing ferritinophagy or lipophagy while transiently restraining mitophagy can flip net autophagy from a survival gate to a pro-ferroptotic accelerator. Consistent with our lipid-focused sections, the SNORD88C–SCD1 axis provides an epitranscriptomic lever over membrane composition: elevated SCD1 skews MUFA/PUFA balance and suppresses both autophagy and lipid peroxidation, predicting reduced ferroptosis sensitivity in SNORD88C-high tumors [61].

A falsifiable workflow can anchor this model in lung cancer: (1) quantify labile iron (e.g., calcein quenching) and lipid ROS (C11-BODIPY) alongside NCOA4–FTH1 colocalization and autophagy flux (LC3-II/p62 under short lysosomal blockade; mCherry-GFP-LC3) in NSCLC lines, organoids, and biopsies [94]; (2) establish causality using NCOA4 loss/gain and FTH1/FTL stabilization, with ferroptosis rescue by ferrostatin-1/liproxstatin-1; (3) modulate lipid circuitry (ACSL4/LPCAT3 versus SCD1 inhibition or MUFA supplementation) to prove substrate dependency [95]; (4) manipulate mitophagy (e.g., USP30 inhibition or Atg7 perturbation) while monitoring mitoROS (MitoSOX), membrane potential, and oxygen consumption; and (5) validate in KL-biased PDX/GEMMs with paired on-treatment sampling. Stratification by KEAP1/NRF2 status and hypoxia scores is advisable to control confounding antioxidant programs.

Therapeutically, combining ferroptosis induction with selective autophagy modulation (enhancing ferritinophagy or preventing compensatory mitophagy) offers a route to flip autophagy from a survival shield to a lethal catalyst. Practically, pulse-timed regimens—initiating lipid peroxidation (e.g., system Xc^– or GPX4 blockade) followed by short-window ferritinophagy enhancement and/or mitophagy restraint—may maximize tumor-selective killing while preserving immune and normal-tissue autophagy. Biomarkers such as labile iron load, NCOA4 puncta density, lipid-ROS kinetics, and flux readouts should be embedded as PD endpoints to de-risk translation.

While candidates such as SNORD88C, circRAPGEF5, and retinoblastoma binding protein 4 (RBBP4) show promise, their clinical utility remains unproven. To avoid overstatement, we explicitly distinguish the evidentiary ladder: analytical validity, clinical validity, clinical utility. Robust validation requires prespecified cut-offs, multi-center prospective cohorts, and head-to-head comparison against established benchmarks (e.g., PD-L1, ctDNA-MRD), ideally with decision-curve or net-benefit analyses to demonstrate added value. Pre-analytic variables (specimen type, ischemic time), intra-tumoral heterogeneity, and longitudinal dynamics must be controlled if these markers are to guide autophagy-targeted therapy or trial enrollment. Accordingly, we downgrade the language from “biomarker” to “candidate biomarker” where appropriate and outline study designs that can move these candidates along the validation pathway.

SNORD88C, which enhances SCD1 translation and suppresses autophagy/lipid peroxidation, should have analytical validity demonstrated by cross-platform quantification in tissue and biofluids (small-RNA sequencing, RT-qPCR, and ddPCR) with spike-in normalization, hemolysis controls (e.g., miR-451/miR-23a index), freeze–thaw stability, and inter-laboratory reproducibility [61]. Clinical validity requires multi-center cohorts relating SNORD88C levels to diagnosis (NSCLC vs benign), prognosis (OS/PFS), and therapy response, with head-to-head comparison against PD-L1/TMB/ctDNA-MRD and optimism-corrected AUC, calibration, and DeLong tests. Clinical utility should be tested in biomarker-stratified trials where SNORD88C-high tumors—hypothesized to be MUFA-skewed and autophagy-suppressed—are randomized to standard care $\pm$ SCD1 inhibition and/or autophagy-rebalancing strategies, with net benefit assessed by decision-curve analysis.

circRAPGEF5, which inhibits autophagosome–lysosome fusion via an m6A-dependent “mRNA trap” (IGF2BP2–NUP160), analytical validity demands RNase-R resistance, junction-spanning assays, orthogonal confirmation by Sanger sequencing, and MeRIP-qPCR/RIP to prove m6A/IGF2BP2 engagement; subcellular fractionation should confirm cytoplasmic enrichment relevant to autophagy trafficking [96]. Clinical validity involves correlating circRAPGEF5 with flux phenotypes (LC3-II/p62 under short CQ/bafilomycin pulse; mCherry-GFP-LC3 reporter where available), metastatic risk, and therapeutic outcomes. Clinical utility could be addressed in an adaptive-strategy design testing whether circRAPGEF5-high tumors benefit from agents that restore autophagic flux (e.g., lysosome re-acidification or SNARE/ESCRT facilitation) combined with EGFR-TKIs or chemotherapy, with pre-specified interaction testing.

RBBP4, a chromatin regulator whose loss induces autophagy-dependent cell death, analytical validity requires standardized immunohistochemistry with digital H-score, intra/inter-observer concordance, and concordant protein/RNA measurements; HDAC1/2 activity assays can provide mechanistic orthogonality. Clinical validity should evaluate associations between RBBP4 levels, autophagy markers (LC3-II/Beclin-1/p62), and outcomes in independent cohorts [97]. Clinical utility is best tested in a biomarker-stratified trial where RBBP4-low (autophagy-elevated) patients receive pulse-timed autophagy inhibition plus chemotherapy or targeted therapy, with predefined PD endpoints (tumor LC3-II/p62 dynamics) to confirm on-target modulation.

To ensure rigor across markers, we recommend: (i) pre-registered statistical analysis plans with pre-specified thresholds (avoiding data-driven cut-offs); (ii) internal cross-validation and external validation with calibration, net reclassification improvement (NRI) and integrated discrimination improvement (IDI); (iii) systematic control of pre-analytic variables (fixation/ischemic time, tube type, processing timelines, freeze–thaw cycles); (iv) embedding autophagy flux readouts (LC3-II/p62 under short lysosomal blockade; mCherry-GFP-LC3) as PD companions; and (v) genomic/metabolic stratification (e.g., KRAS/LKB1, KEAP1/NRF2, TP53, hypoxia scores) to mitigate confounding. Taken together, these steps operationalize the analytical$\to$clinical validity$\to$clinical utility ladder and provide a concrete path for SNORD88C, circRAPGEF5, and RBBP4 to transition from candidate markers to clinically actionable tools—if and only if they demonstrate incremental value over current standards in prospective, multi-center studies.

RNA epigenetic modifications have recently emerged as a critical layer of autophagy regulation, with significant implications for translational research in lung cancer therapy. One of the most studied modifications is N6-methyladenosine (m6A) on mRNA, which can influence mRNA stability and translation. Studies have shown that m6A modifications modulate the expression of autophagy-related molecules, thereby affecting chemosensitivity and metastatic progression in lung cancer [129]. In NSCLC, METTL3 (an m6A methyltransferase) was found to drive abnormal autophagy, particularly mitophagy, through two mechanisms [108]: first, METTL3-mediated m6A led to the accelerated degradation of DCP2 mRNA, resulting in continuous activation of the PINK1-Parkin mitophagy pathway; second, METTL3 enhanced the nuclear translocation of the master autophagy/lysosome regulator TFEB, promoting the transcription of autophagy-lysosome genes. Together, these effects caused excessive mitophagy, which contributed to the development of cisplatin resistance. Encouragingly, a selective METTL3 inhibitor (STM2457) was able to restore cisplatin sensitivity in resistant cells and also reprogram the tumor microenvironment in vivo to enhance the efficacy of PD-1 checkpoint blockade. This indicates that targeting RNA methylation can have multifaceted therapeutic benefits, both direct (tumor-intrinsic) and indirect (immune-mediated).

On the flip side of autophagy regulation, recent research has uncovered an intricate co-regulation network involving RNA modifications and the ubiquitin-proteasome system. Homeobox containing 1 (HMBOX1) was shown to influence autophagy homeostasis via an m6A-dependent mechanism [130]. Specifically, METTL3-catalyzed m6A modification increased the stability of HMBOX1 mRNA, leading to upregulated HMBOX1 protein, which in turn enhanced the expression of the E3 ubiquitin ligase HACE1. HACE1 was found to mediate K63-linked ubiquitination of ATG5, tagging it for proteasomal degradation and thus acting as a brake on autophagy. The HMBOX1-HACE1-ATG5 axis appears to fine-tune autophagy levels, and it has implications for chemosensitivity—elevated HMBOX1/HACE1 activity correlates with reduced autophagy and potentially greater sensitivity to certain drugs like 5-fluorouracil (5-FU). These findings suggest that components of the RNA epitranscriptomic machinery can serve as biomarkers for optimizing chemotherapy, and that targeted disruption of specific RNA modification readers or writers could recalibrate autophagy in cancer cells.

Circular RNAs (circRNAs) add another layer to the regulatory landscape. One example is circRAPGEF5, which was found to inhibit autophagy through an m6A-dependent “mRNA trap” mechanism [28]. m6A-modified circRAPGEF5 can form a complex with the m6A reader IGF2BP2 and the mRNA of NUP160 (a nucleoporin), leading to stabilization of NUP160 mRNA. The increased NUP160 in turn interferes with autophagosome-lysosome fusion, effectively stalling autophagic flux. Functionally, this enhances metastatic potential in lung adenocarcinoma. This insight not only elucidates a novel mechanism underlying metastasis but also suggests that circRNAs themselves (or their m6A status) could be therapeutic targets or diagnostic markers. Collectively, these advancements highlight RNA epigenetic modification networks (including m6A writers like METTL3, erasers like FTO, and readers like IGF2BP2) as promising intervention points to reverse lung cancer drug resistance and inhibit metastasis by modulating autophagy.

Anti-angiogenic therapy has become a cornerstone of cancer treatment, and in lung cancer its role is expanding beyond simply starving tumors of blood supply. There is growing recognition that anti-angiogenic agents can also influence autophagy and tumor immunity. Apatinib is one such agent, a multi-tyrosine kinase inhibitor that primarily targets VEGFR2. In NSCLC, apatinib has been observed to exert multiple anti-tumor effects. It inhibits tumor angiogenesis by blocking the VEGFR2/STAT3 signaling pathway and simultaneously downregulates oncogenic and immunosuppressive proteins like PD-L1 and c-Myc, thereby enhancing anti-tumor immune responses. Crucially, apatinib can induce an accumulation of ROS within tumor cells, which has downstream effects on autophagy. Elevated ROS from apatinib treatment was shown to inhibit the Nrf2/p62 pathway, leading to the promotion of autophagy-dependent cell death (a form of autophagic apoptosis) [130]. In essence, apatinib creates a multi-pronged assault on the tumor: it prunes the vasculature, reactivates immune elements, and drives cancer cells toward autophagy-dependent death. Additionally, Apatinib, in combination with camrelizumab, enhances the efficacy of neoadjuvant immunotherapy for oral squamous cell carcinoma (OSCC) by improving major pathological response (MPR) rates, with higher MPR observed in patients with elevated PD-L1 expression and increased CD4+ T-cell infiltration [109]. This suggests potential utility of apatinib in overcoming resistance mechanisms and immune evasion in lung cancer.

Anti-angiogenic therapy also reshapes autophagy and tumor immunity via a limited set of recurring mechanisms. In brief: VEGFR2 blockade (e.g., apatinib) elevates ROS, suppresses Nrf2/p62 signaling, and drives ADCD while lowering PD-L1/c-Myc [131]; vascular normalization (e.g., Endostar with PD-1 blockade) improves T-cell infiltration and, through PI3K/Akt/mTOR tuning, increases tumor-cell autophagy to enhance ICI efficacy; lysosome-targeting natural products (e.g., baicalin) alkalinize lysosomes and block autophagy flux [110]; and ROS-dependent agents (e.g., halofuginone) co-induce autophagy and apoptosis. To reduce redundancy, repeated assay readouts (LC3-II/p62 dynamics, lysosomal pH changes) are omitted here while preserved in the cited studies [111].

Targeting autophagy at the level of specific organelles has emerged as a novel therapeutic strategy, as different organelle-specific autophagy processes (such as mitophagy, ER-phagy, ribophagy, and others) can be selectively manipulated to kill cancer cells. Several breakthroughs illustrate the potential of organelle-targeted autophagy modulation in lung cancer.

Interactions between autophagy regulation and the immune system are increasingly recognized as crucial in determining the success of cancer immunotherapy. Several key autophagy nodes have been identified that modulate anti-tumor immunity, offering new opportunities to overcome immune resistance in lung cancer.

Nanomedicine has introduced new possibilities for delivering autophagy-targeting therapies with improved precision and efficacy. By using nanoparticles and other nanoscale delivery vehicles, we can overcome challenges such as poor solubility of drugs, off-target effects, and drug resistance mechanisms. Several innovative nano-delivery systems have been developed to specifically modulate autophagy in lung cancer.

Despite compelling preclinical data, clinical trials with CQ/HCQ or mTOR inhibitors in lung cancer have yielded variable efficacy [115]. Convergent explanations include: (i) insufficient tumor exposure and incomplete flux inhibition at tolerated doses; (ii) lack of real-time PD readouts (LC3-II/p62 dynamics with lysosomal blockade, lysosomal pH imaging); (iii) suboptimal timing with chemotherapy/EGFR-TKIs (continuous blockade may dampen immunogenic stress signals); (iv) off-target toxicities constraining dose intensity; (v) failure to stratify by autophagy-addicted genotypes/phenotypes (e.g., KL subtype). Actionable trial design upgrades: embed PD/PK biomarkers, adopt pulse-timed schedules synchronized to cytotoxic peak stress, prioritize lung-targeted delivery (e.g., inhalable CQ analogs), and pre-select patients by autophagy/metabolic signatures to enrich for benefit.

Given the heterogeneous roles and context-dependence of autophagy in lung cancer, clinical use of autophagy modulators should be anchored in three pillars: (i) biomarker-defined patient selection, (ii) prospective safety monitoring tailored to mechanism, and (iii) PD proof of on-target autophagy modulation in tumor tissue. First, selection should enrich for putative “autophagy-addicted” contexts—e.g., KL genotype (KRAS with LKB1 loss), KEAP1/NRF2 alterations, high TFEB/lysosomal signatures, mitophagy-high phenotypes, or elevated basal flux (LC3-II/p62 dynamics under short CQ/bafilomycin pulses). Where feasible, ferritinophagy/mitophagy readouts (NCOA4–ferritin puncta; PINK1/Parkin recruitment; mt-Keima ratio) and lipid circuitry (SCD1/MUFA–PUFA balance) can refine eligibility.

Second, safety profiles are mechanism-specific and warrant proactive monitoring. Lysosomotropic agents (e.g., CQ/HCQ) carry risks of QT prolongation, retinopathy (cumulative dose), cytopenias, and GI intolerance; rapalogs/mTOR inhibitors may cause stomatitis, hyperglycemia/dyslipidemia, and non-infectious pneumonitis; broad or prolonged autophagy suppression can impair immune and muscle/organelle homeostasis. Practical mitigations include pulse-timed/sequence-based dosing (rather than continuous exposure), ECG monitoring when combined with TKIs, baseline ophthalmologic assessment for long courses, metabolic panels (glucose/lipids), and parallel immune fitness checks (e.g., LC3/ATG signatures in T/NK subsets).

Third, trial designs should mandate in-tumor PD confirmation (e.g., TFEB nuclear translocation, LC3-II kinetics under short BafA1/CQ pulses, lysosomal pH imaging) as co-primary or key secondary endpoints, with pre-specified stop/go rules tied to flux correction. Biomarker-embedded, enrichment or adaptive randomization strategies are preferred, and tumor-preferential delivery (inhalable or nano-enabled formulations; organelle-targeted agents) can widen the therapeutic window.

Current and prior clinical attempts (e.g., CQ/HCQ combinations; rapalog-based strategies) have yielded mixed efficacy—likely reflecting incomplete intratumoral flux blockade at tolerated doses, suboptimal scheduling relative to therapy-induced stress, and lack of PD/selection. We therefore advocate biomarker-guided, pulse-timed regimens that synchronize autophagy modulation with chemotherapy, EGFR-TKIs, or immune checkpoint inhibitors, while embedding safety and PD gates to balance antitumor benefit against on-target liabilities.

The integration of autophagy-targeted strategies with standard lung cancer treatments has shown great promise in overcoming resistance and improving efficacy. Below we outline how autophagy modulation can be combined with chemotherapy, targeted therapy, immunotherapy, and other novel strategies, highlighting recent advances and future prospects:

Autophagy supports EMT by supplying acetyl-CoA for chromatin remodeling (e.g., Snail acetylation) and by buffering redox stress during invasion; conversely, excessive selective autophagy of focal-adhesion/$\beta$-catenin components can restrain motility—implying a non-linear relationship that may be exploited to time autophagy modulation around metastatic bottlenecks [69]. Mechanistically, autophagy-derived acetyl-CoA sustains CBP/p300-dependent acetylation of EMT TFs (e.g., Snail), while mitophagy curbs mitoROS to permit survival during detachment and intravasation; in contrast, p62/SQSTM1-mediated targeting of $\beta$-catenin or adhesion scaffolds can limit protrusion dynamics and migration [161]. Operationally, we propose pulse-timed autophagy modulation at two bottlenecks—(i) early dissemination (to blunt survival during anoikis and circulation) and (ii) post-extravasation seeding (to prevent metabolic adaptation)—guided by PD readouts such as LC3-II/p62 dynamics under short CQ/BafA1 pulses, Snail acetylation status, and ¹³C-tracing of autophagy-derived acetyl-CoA [162]. Lipid circuitry intersects this axis: SCD1/MUFA accumulation lowers lipid peroxidation and may reduce EMT-coupled autophagy stress, suggesting that SCD1 inhibition combined with autophagy re-balancing could constrain invasion in MUFA-skewed tumors.

SCLC exhibits high basal autophagy/mitophagy linked to chemo-resistance. Targeting mitophagy regulators (e.g., METTL3-TFEB axes or BEX2-NDP52 interactions) can resensitize cells to platinum/etoposide in models, warranting SCLC-tailored trials with mitophagy-centric PD endpoints [108]. Given the neuroendocrine lineage and rapid cycling of SCLC, mitophagy sustains mitochondrial quality and ATP homeostasis during platinum-induced stress; enforcing mitophagy blockade unmasks ROS-driven death. Prospective SCLC studies should embed mitophagy PD (mt-Keima ratio, PINK1/Parkin recruitment, TFEB nuclear translocation) alongside clinical endpoints, and test pulse-timed mitophagy restraint (e.g., USP30 inhibition or late-stage fusion blockade) layered onto platinum/etoposide, with mandatory tumor or ctPD surrogates confirming on-target flux modulation. Because autophagy also shapes antigen processing, combinations with PD-1/PD-L1 blockade merit evaluation only when PD shows tumor-selective autophagy engagement, to avoid compromising T/NK cell fitness.

Early evidence suggests that intratumoral and airway microbiota can tune epithelial and myeloid autophagy, thereby shaping antigen presentation and checkpoint responses. Defining causality—and whether selective autophagy modulation can “normalize” dysbiotic signals—represents a tractable, high-yield research direction. Microbial ligands (e.g., LPS, peptidoglycan) engage TLR/cGAS–STING pathways that intersect with autophagy/xenophagy [163], influencing PD-L1 turnover, MHC-I presentation, and cytokine milieus. We advocate paired airway/tumor microbiome profiling (16S/shotgun plus metabolomics) with epithelial/myeloid autophagy PD (LC3 puncta, p62 turnover, lysosomal pH) to map microbe–autophagy–immunity triads and identify actionable “dysbiosis-autophagy” phenotypes. Interventional concepts—such as short-course antibiotics avoidance, pre-/post-biotics, or bacterially derived autophagy modulators—should be tested cautiously in window-of-opportunity designs, with predefined safety gates and immune monitoring to preserve antitumor immunity.