The role of autophagy in ocular health: mechanisms, pathologies, and therapeutic strategies Authors: Mingyu Huang, Yumeng Lin, Hong Zhang, Ruimin Yuan, Liuzhi Zeng, Sizhen Li, Qingsong Yang, Meng Gong, Renyan Xiao, Peng Jia, Lan Yuan, Qian Guo, Zhongyu Han, Lidan Xie Published: 2026 ABSTRACT Autophagy is a self-digestive process in which cellular components are degraded and recycled to maintain homeostasis and cope with stress. When cells are in a state of stress, autophagy will degrade proteins, lipids and damaged organelles, and convert them into key nutrients and building cornerstones needed to maintain cell homeosis. More and more evidence shows that there is a link between autophagy process damage and the development of a variety of ophthalmic diseases that seriously threaten vision, including age-related macular degeneration, glaucoma and the formation of cataracts. Through a systematic review, we integrate experimental data covering basic biological mechanisms, genetic models and clinical evidence to clarify these pathological associations. With a focus on the mechanisms and regulation of autophagy in various ocular tissues and diseases, we analyzed research on autophagy in the lens, retina, cornea, and trabecular meshwork. Key findings elucidate the specific mechanisms of autophagy in maintaining lens transparency, promoting retinal ganglion cell survival, and regulating ocular immunity, demonstrating its critical role in preserving cellular equilibrium in ocular tissues. Furthermore, we evaluated potential therapeutic strategies targeting autophagic pathways, including mammalian target of rapamycin (mTOR) inhibition, transcription factor EB(TFEB) activation, noncoding RNA regulation, gene editing, and artificial intelligence-assisted diagnostics, which show significant promise for modulating autophagy to treat ocular diseases. The results of this review underscore the importance of autophagy in ocular health and disease. Introduction Ocular disorders significantly impact global health, with age-related macular degeneration (AMD), diabetic retinopathy (DR), glaucoma, and cataracts being the main causes of irreversible vision loss. AMD is the leading cause of vision loss in individuals over 50 years of age and affects millions of people globally [1]. Up to 80% of individuals with long-standing diabetes are affected by DR, with severe cases potentially requiring laser therapy or surgery [2]. Glaucoma pathophysiology typically features abnormal intraocular pressure elevations that trigger progressive retinal ganglion cell degeneration and functional visual deficits, establishing this disease group as the second leading source of blindness internationally [3]. Cataracts, characterized by blurred vision, increased sensitivity to light, and faded colors, have an incidence that significantly increases with age [4]. Autophagy is a self-degradative process for intracellular component recycling. Three types are distinguished by lysosomal delivery mechanisms. The macroautophagy pathway initiates with the encapsulation of cytoplasmic constituents within a bilayered membrane structure that matures into an autophagosome. The vesicle then fuses with the lysosomal chamber and realizes enzymatic degradation of its internal substances with the help of acid hydrolase [5]. Partner-mediated autophagy (CMA) is a selective degradation pathway unique to mammalian cells. The process involves Hsc70 companion protein identification of target proteins with specific KFERQ amino acid sequences [6]. Microautophagy involves the entry of degradable materials into lysosomes via invaginations of the lysosomal or endosomal membrane. This review centers on macroautophagy, subsequently termed autophagy [7]. As a fundamental cellular process, autophagy operates continuously across biological systems to support macromolecular turnover, energetic equilibrium, and regulatory signaling [8]. This process is upregulated by stressors including nutrient deprivation, hypoxia, and infection, wherein it degrades cellular constituents to sustain cellular metabolism [9]. Critically, this pathway functions as a selective quality control system through organellophagy and clearance of aberrant proteins. Autophagic dysfunction consequently perturbs physiological balance, establishing pathogenic foundations for malignancies, neurological disorders, and immune pathologies [8]. This review synthesizes investigations conducted since the name “autophagy” was proposed, spanning its multifaceted roles in ocular health and disease (Fig. 1). The assessment emphasizes its fundamental capacity to preserve cellular equilibrium while demonstrating substantial participation across multiple disease mechanisms. In ocular systems, autophagy contributes critically to tissue development, homeostatic maintenance, and degenerative pathways, affecting transparent media, the neural retina, anterior segment structures, and aqueous humor regulation. Autophagy dysregulation represents a pathogenic factor in multiple blinding conditions. This review offers a detailed overview autophagy in ocular diseases and its potential as a therapeutic target. Fig. 1Timeline of pivotal autophagy discoveries and their ocular tissue applications. This timeline outlines key discoveries in autophagy and their relevance to ocular biology. Since Christian de Duve coined the term in 1963, landmark studies have elucidated its mechanisms, advancing from yeast and cellular models to mice, and ultimately revealing its critical roles in retinal homeostasis and pathological processes. The selection is not exhaustive due to space constraints. Atg, autophagy-related gene; CMA, chaperone-mediated autophagy; PI3K, phosphoinositide 3-kinase; p62/SQSTM1, sequestosome 1; RPE, retinal pigment epithelium; AMD, age-related macular degeneration; ONT, optic nerve transection; RGC, retinal ganglion cell; DR, diabetic retinopathy; RB1CC1, rb1-inducible coiled-coil 1; cSLO, confocal laser scanning ophthalmoscopy Timeline of pivotal autophagy discoveries and their ocular tissue applications. This timeline outlines key discoveries in autophagy and their relevance to ocular biology. Since Christian de Duve coined the term in 1963, landmark studies have elucidated its mechanisms, advancing from yeast and cellular models to mice, and ultimately revealing its critical roles in retinal homeostasis and pathological processes. The selection is not exhaustive due to space constraints. Atg, autophagy-related gene; CMA, chaperone-mediated autophagy; PI3K, phosphoinositide 3-kinase; p62/SQSTM1, sequestosome 1; RPE, retinal pigment epithelium; AMD, age-related macular degeneration; ONT, optic nerve transection; RGC, retinal ganglion cell; DR, diabetic retinopathy; RB1CC1, rb1-inducible coiled-coil 1; cSLO, confocal laser scanning ophthalmoscopy Overview of autophagy Autophagy is a highly conserved degradation mechanism in phylogeny, during which cytoplasmic substances are encapsulated in double-layer membrane vesicles. These autophagosomes then fuse with lysosomal compartments, thereby achieving the decomposition and reuse of their contents [10]. The concept of autophagic flux encompasses this entire progression from vesicle biogenesis through lysosomal fusion and ultimately cargo processing, which provides a basic indicator for evaluating autophagy function [11]. The next chapter will elaborate on the core mechanism of autophagy and the key biomarkers that regulate autophagic flux (Fig. 2). Fig. 2Autophagy core regulatory pathway. The diagram outlines macroautophagy, a sequential pathway comprising initiation, phagophore nucleation, vesicle expansion, lysosomal fusion, and cargo breakdown. Key regulatory modules include the ULK1 initiation complex, PI3KⅢnucleation complex, ATG12 conjugation system and LC3 conjugation systems. Atg, autophagy-related gene; ULK1, unc-51-like autophagy activating kinase 1; FIP200, FAK family-interacting protein of 200 kDa; PI3KⅢ, class III phosphoinositide 3-kinase; VPS34, vacuolar protein sorting 34; VPS15, vacuolar protein sorting 15; LC3, microtubule-associated protein 1 A/1B-light chain 3; mTOR, mechanistic target of rapamycin; AMPK, AMP-activated protein kinase; Beclin1, Beclin1; WIPls, WD-repeat domain phosphoinositide-interacting proteins; DFCP1, double FYVE-containing protein 1; ER, endoplasmic reticulum Autophagy core regulatory pathway. The diagram outlines macroautophagy, a sequential pathway comprising initiation, phagophore nucleation, vesicle expansion, lysosomal fusion, and cargo breakdown. Key regulatory modules include the ULK1 initiation complex, PI3KⅢnucleation complex, ATG12 conjugation system and LC3 conjugation systems. Atg, autophagy-related gene; ULK1, unc-51-like autophagy activating kinase 1; FIP200, FAK family-interacting protein of 200 kDa; PI3KⅢ, class III phosphoinositide 3-kinase; VPS34, vacuolar protein sorting 34; VPS15, vacuolar protein sorting 15; LC3, microtubule-associated protein 1 A/1B-light chain 3; mTOR, mechanistic target of rapamycin; AMPK, AMP-activated protein kinase; Beclin1, Beclin1; WIPls, WD-repeat domain phosphoinositide-interacting proteins; DFCP1, double FYVE-containing protein 1; ER, endoplasmic reticulum Initiation of autophagy The initiation of autophagy is a tightly orchestrated cellular process triggered by diverse intracellular stressors, including nutrient deprivation, energy depletion (ATP deficit), oxidative imbalance, and organellar damage. At the core of this regulatory phase lies the interplay between AMP-activated protein kinase (AMPK) and mechanistic target of rapamycin complex 1 (mTORC1). AMPK activation counteracts mTORC1-driven anabolic repression, thereby synchronizing metabolic reprogramming with autophagic induction [12, 13]. When nutrients are abundant, mTORC1 inhibits the formation of autophagosomes by phosphorylating and deactivating the ULK1 kinase complex, which includes ULK1, ATG13, FIP200, and ATG101 [14]. ULK1, a conserved Ser/Thr kinase, is essential for autophagy initiation. During energy stress (e.g., ATP depletion), AMPK activation induces ULK1 phosphorylation at regulatory sites (Ser317/Ser777), antagonizing mTORC1-mediated inhibition and facilitating ULK1 complex oligomerization [15]. Structural analyses show that the ULK1 complex adopts a supramolecular scaffold configuration, with its C-terminal EAT domain anchoring vesicles and ATG13’s intrinsically disordered regions (IDRs) promoting macromolecular organization at the phagophore assembly site (PAS) [16]. In addition to phosphorylation, the ULK1 complex is also subject to regulation by other post-translational modifications. For instance, the LUBAC complex catalyzes the linear ubiquitination of proteins such as ATG13, while the deubiquitinase OTULIN can reverse this modification; together, they coordinately regulate the stability of the ULK1 complex and the progression of autophagy initiation [17]. Beyond the core initiation machinery governing non-selective autophagy, selective autophagy represents a refined mechanism with distinct regulatory features. Damaged mitochondria, invading pathogens, and aggregated proteins, among other cargoes, are tagged with ubiquitin chains (e.g., K63, K48, K27 linkages) by specific E3 ubiquitin ligases. These ubiquitinated signals are subsequently recognized by selective autophagy receptors (e.g., OPTN, NDP52, p62) or adaptor proteins, which in turn initiate the autophagic process [18]. It is also worth noting that the above regulatory principles may show differences in cell type specificity. For example, the accumulation of falsely folded erytheria is an important cause of autosomal dominant retinal pigmentation (adRP). This accumulated protein can act as a continuous trigger to activate the autophagy initiatory pathway, which may lead to autophagy system overload and subsequent failure [19]. This mechanism is essential for elucidating the pathogenesis of related diseases. Formation of autophagosomes Upon the induction of autophagy, a partial bilayer membrane emerges as the phagophore. This forming sac-like structure then extends and evolves into mature autophagy vesicles. The biological development of the organelle involves a progressive membrane remodeling process, that is, through the continuous fusion of the membrane source until the complete closure of the vesicle [20]. After the ULK complex is activated, a specific III phosphatidylinositol 3-kinase complex (PI3KC3-CI/II) containing BECN1 is assembled and formed to catalyze the generation of phosphatidylinositol 3-phosphate (PtdIns3P). This lipid second messenger then promotes the initiation of the phagocytic membrane and promotes the maturity of the endome-lysosomal compartment and autophagy lysosome [21]. Recent work identifies Transferrin Receptor as critical for autophagosome formation and closures [22]. On RAB11A-positive recycling endosomal membranes, the ubiquitin ligase MARCH8 catalyzes TfR ubiquitination. This modified TfR acquires the capacity to recruit VPS15, an essential subunit of the VPS34 complex I, thus promoting PI(3)P production. The generated PI(3)P then attracts effector proteins including WIPI2, initiating autophagosome biogenesis [23]. The biological occurrence of autophagy membranes is still an unresolved core problem in the field of autophagy research. Vesicles containing ATG9 act as key membrane scaffolds in the process of phagocytic nucleation; they play the role of lipid carriers and establish a connection between the endoplasmic reticum (ER) and the new phagocytocy through the latation mechanism mediated by ATG2A [24, 25]. The elongation of the phagophore, marked by developing bilayer membranes, is principally orchestrated by the Atg6/Beclin1 complex through the modulation of class III phosphatidylinositol 3-kinase (PI3KC3) function. Membrane contact sites (MCSs), especially those between the endoplasmic reticulum (ER) and plasma membrane, supply additional lipid resources during autophagosome formation [26]. Further experimental data show that in the microenvironment of hepatocellular carcinoma, a metabolite produced by adenosine succinate lyase (ADSL), can inhibit lysine demethylase 8 (KDM8) [27]. This inhibitory effect will increase the methylation level of Beclin1, thus enhancing the catalytic activity of PI3KC3 and promoting the activation of malignant intracellular autophagy. However, the environmental background of the above mechanisms is highly specific. Therefore, the correlation between these mechanisms and basic autophagy in the tissues of differentiated mammals in the body cannot be clearly established at present. In view of this, it is still necessary to carry out more extensive experimental verification work. Elongation of autophagosomes The phagocytic vesicle extends its membrane through de novo phospholipid synthesis. In yeast and mammals, long-chain acyl coenzyme A synthase (such as Faa1 and ACSL4) plays a key role in catalyzing fatty acid activation, thus driving phospholipid synthesis and providing lipid sources for autophagosome membranes [28]. GRAMD1A, functioning as a cholesterol transporter, indirectly regulates the elongation process of the autophagosome membrane by supplying cholesterol to phagosomes. Conversely, GRAMD1C exerts a constraining effect on autophagosome biogenesis and attenuates mitochondrial bioenergetic output [29]. Governing essential sterol exchange at key membrane contact sites, this mechanism facilitates cholesterol trafficking between the plasma membrane, endoplasmic reticulum, and mitochondria. The extension stage requires two ubiquitin-like coupling systems involving Atg12 and LC3 modifications. Atg12-Atg5-Atg16 polymers are mainly located on the cytoplasmic surface of the forming autophagosome. Moreover, lipidated Atg8/LC3 decorates both sides of the autophagosomal membrane and facilitates bilayer expansion of phagophores during autophagosome maturation [30]. At the same time, as the only transmembrane ATG protein, ATG9A provides membrane substances through vesicle fusion with the new autophagy structure, thus driving the growth of the membrane [31]. These vesicles will dynamically fuse with the phagocytic vesicles, thus promoting the extension of the membrane. Contemporary studies have clarified how the membrane contact site is formed between the expanding phagophore and the endoplasmic reticulum. The TRAPPIII complex, specifically via its Trs85 subunit, collaborates with Atg2 to establish these interorganellar junctions. These connections enable TRAPPIII-mediated activation of the Rab GTPase Ypt1 (functionally analogous to mammalian RAB1), which subsequently stimulates local class III PI3-kinase complex I (PI3KC3-C1) to generate phosphatidylinositol 3-phosphate [32]. This lipid messenger then recruits downstream effectors including Atg18/Atg21, coordinating membrane expansion through lipid transfer and structural remodeling [33]. Notably, different stages and regions of autophages may involve completely different membrane lipid sources and molecular mechanisms. At the mechanical level, how the two are coupled is still an outstanding issue and deserves further in-depth exploration. Maturity of autophagosomes Autophagosomal development is a dynamically regulated process that involves the synchronized activities of multiple molecular assemblies, with membrane fusion representing its central mechanistic element. Posttranslational modifications such as phosphorylation and ubiquitination act as critical modulators that govern the functional capacity, protein‒protein interactions, and compartment-specific localization of the principal factors driving autophagosomal maturation. In addition, the types of phosphoinositide lipids on the membrane (such as PtdIns3P) and their spatial distribution provide the necessary platform and identification signals for the recruitment of RAB proteins, tethering factors and SNARE proteins, thus constituting a key mechanism for the precise regulation of membrane fusion [34, 35]. Once cargo is completely wrapped, the phagophore seals will close the edge of its membrane, thus transforming into a closed double-layer autophagy. ESCRT-I components such as VPS28 mediate the assembly of ESCRT-III elements including CHMP2A and CHMP4B at the phagophore’s periphery. Subsequently, the polymerized ESCRT-III filamentary structure mediates membrane contraction, while AAA+ ATPase VPS4 catalyzes the depolymerization of the complex, thus achieving the final closure [36]. Neurodegeneration studies confirm that impaired ESCRT-III recruitment (e.g., via TfR-4KA mutation) causes accumulation of unsealed phagophores, promoting Aβ aggregation [37]. Notably, certain viruses exploit this machinery for survival. For instance, classical swine fever virus (CSFV) nonstructural proteins (NS4B and NS5A) bind host ESCRT components (CHMP1A/CHMP4B), hijacking phagophore closure to accelerate autophagic flux and facilitate viral replication [38]. In the process of phagocytic closure, TfR-mediated phosphatidylinositol 3-phosphate production enables recruitment of ESCRT pathway elements including VPS36. Subsequently, this molecular machine drives the expanding phagocytic bubble to complete the final closure, thus transforming it into a mature autophagosome [22]. Contemporary investigations demonstrate that cells deploy an alternative Rab27a-regulated secretory pathway derived from autophagy when confronting obstacles in endolysosomal function or autophagosomal maturation, such as those imposed by the SARS-CoV-2 ORF3a viral protein. Autophagy cargo receptors such as p62/SQSTM1 are packaged into extracellular vesicles and particles (EVPs) and released outside the cell [39]. Autophagosome-lysosome fusion Upon maturation, autophagosomes fuse with lysosomal compartments to form autolysosomes, which subsequently degrade sequestered contents through hydrolytic enzymes [40]. Newly formed autophagosomes do not remain stationary; instead, they are actively transported to the perinuclear region, where lysosomes are concentrated, via microtubules and motor protein complexes [41]. To facilitate the fusion of autophagosome and lysosome membranes and promote the formation of SNARE complexes, various membrane-anchoring factors play a crucial role. For example, the heterohexameric HOPS (homotypic fusion and protein sorting) complex tethers autophagosomes to lysosomes and promotes SNARE complex assembly by interacting with Syntaxin 17 (STX17) to drive membrane fusion [42]. The latest research findings indicate that deacetylated SNAP47 can directly recruit the HOPS complex to autophagosomes. This process is independent of STX17 and the formation of the SNARE complex, while acetylated SNAP47 inhibits this recruitment [43]. PLEKHM1 serves as a multivalent adaptor protein that bridges autophagic and endocytic trafficking pathways by binding to the HOPS complex via its Rubicon homology domain, interacting with LC3 through its LIR motif, and localizing to late endosomes/lysosomes in a RAB7-dependent manner [44]. Its dysregulation leads to impaired autophagic flux [45]. Following tethering, STX17 on the autophagosome membrane forms a trans-SNARE complex with VAMP8 and SNAP29 on the lysosome membrane, driving the fusion of the double-layer membranes and transporting the autophagic contents to the lysosomal cavity [46, 47]. EPG5, functioning as a Rab7 effector, acts as a bifunctional tether that concurrently engages LC3 on autophagosomal membranes and VAMP7/8 within lysosomal compartments. Reconstituted membrane fusion assays have established that EPG5 mediates supramolecular assembly of the STX17-SNAP29-VAMP8 ternary complex [48]. Accordingly, impaired EPG5 function results in aberrant buildup of stalled autophagosome-lysosome fusion intermediates. Degradation and autophagic lysosome reformation Ultimately, the enzymes within the autophagosome degrade the proteins, organelles, and other components encapsulated within the autophagosome. Following lysosomal hydrolysis, liberated macromolecules undergo transporter-mediated efflux into the cytosol for subsequent anabolic utilization. Once the autophagosome is formed, lysosomal degradation of the autophagosed material commences immediately. Studies have shown that the ATG-conjugated system promotes the degradation of the inner membrane of the autophagosome whilst maintaining the integrity of the outer membrane [46]. This may be due to the two membranes acquiring different properties during the lysis process. During autophagic degradation, the autolysosome relies on vacuolar-type H⁺-ATPase (V-ATPase) to generate an acidic lumenal environment (pH 4.5–5.0), a prerequisite for activating hydrolytic enzymes [49]. Within this compartment, the sequestered cargo undergoes coordinated breakdown by various hydrolases, including cathepsins (B, D, L) responsible for cleaving proteins into amino acids [50], lysosomal acid lipase (LAL) that hydrolyzes cholesteryl esters and triglycerides [51], DNases and RNases that degrade nucleic acids, and glycosidases that process glycoconjugates [52]. Within the autophagosome, small molecules such as amino acids and fatty acids generated by degradation are recycled into the cytoplasm via specific transporters, where they are reused for anabolic processes and energy production to maintain cellular homeostasis [53]. The efficiency of this degradation and recovery stage directly determines the autophagy rate and the adaptability of cells. The dysfunction of this process will lead to the accumulation of non-degraded products, which constitutes a common pathological basis for diseases such as neurodegenerative diseases and lysosomal storage disease [54]. Autophagic lysosome reformation (ALR) represents the final stage of autophagy. Its function is to regenerate the new lysosomes, thus maintaining the homeosome of the lysomes [55]. This process is triggered by the reactivation of mTOR and acts as a negative feedback mechanism to prevent excessive autophagy. During ALR, autolysosomal membranes bud to form tubular structures that give rise to proto-lysosomes characterized by neutral PH and absence of intraluminal proteins. Subsequently, these pro-lysosomes undergo acidification and obtain lysosomal proteins, thus maturing into functional nascent lysosomes [56, 57]. Methods for investigating autophagy in ocular systems As research on autophagy in ocular diseases has increased, multiple methods have been employed to monitor and modulate autophagic activity, thereby elucidating its role in pathological ocular processes. Transmission electron microscopy (TEM) enables direct visualization of the characteristic double-membrane structures of autophagosomes [58]. Alternatively, fluorescence microscopy utilizing fluorescently tagged LC3 (e.g., GFP-LC3) allows real-time visualization of autophagosome formation and spatial distribution within cellular contexts [59]. However, given the short life of autophagosomes (about 10 min), the number of autophagosomes alone cannot accurately reflect the activity of autophagy [60]. Instead, the rate of autophagic flux is usually evaluated by monitoring the degradation efficiency of autophagy substrates [61]. Specifically, this can be achieved by comparing the accumulation of autophagy markers (e.g., LC3-II) before and after treatment with inhibitors (such as chloroquine or leupeptin), or by monitoring the degradation levels of autophagy-specific substrates like p62 [62]. However, the transcription of p62 is regulated by multiple stress pathways (such as Nrf2), which indicates that its accumulation may suggest that autophagy function is impaired, but the reduction in its level does not necessarily mean that autophagy function is enhanced. The study of gene and protein expression aids in the rigorous assessment of autophagic flux. By quantitatively measuring the levels of autophagy-related transcripts (e.g., ATG5, ATG7, LC3) [62] and combining this with immunoblot analysis of corresponding protein expression and modification states, researchers can obtain multidimensional autophagy parameters. Non-invasive imaging technology makes it possible to visualize and quantitatively analyze autophagy activity in living organisms. The changes in retinal structure that can be detected by optical coherence tomography (OCT) reflect the cell dynamics related to autophagy [63]. For example, the thinning of the retinal nerve fiber layer (RNFL) and ganglion cell complex layer (GCC) measured in the glaucoma model directly corresponds to the loss of retinal ganglion cells (RGC) [64]. Furthermore, hyperreflective foci (HRF) observed in OCT images may correspond to aggregates of abnormal proteins, apoptotic bodies, or degenerated cellular debris. The insufficiency of autophagic clearance mechanisms causes material deposition that enhances HRF visibility in OCT scans [65]. The administration of bafilomycin to normally nourished rodents resulted in increased LC3-II accumulation compared with that in their untreated counterparts, indicating that bafilomycin exhibits constitutive autophagic activity in retinal tissue. Fluorescence imaging of GFP-LC3-labeled retinal sections revealed increased autophagic vesicle formation within the ganglion cell layer, inner nuclear layer, and photoreceptor-enriched outer nuclear layer during nutritional deprivation [66]. Isolated retinal explants allow rapid quantification of autophagic flux across embryonic and mature tissue by Western blot after addition of lysosomal blockers to the culture medium. It is worth noting that the embryonic retina also has another advantage, which is to support whole-eye explant culture; this allows researchers to conduct autophagy evaluation through retinal sections, so as to visually observe LC3 puncta in specific cell groups in the retina. Quantitative evaluation of autophagic activity can additionally employ flow cytometric monitoring of fluorescence alterations in autophagy markers (such as LC3-II and p62), accompanied by lysosomal blockade to examine flux behavior [66, 67]. Autophagy in ocular development and homeostatic maintenance Autophagy pathways play a fundamental role in eye physiology, which not only constitutes an important mechanism for maintaining tissue integrity and cell function in various areas of the eye, but also gives the eye the ability to adapt to environmental challenges. The following sections elaborate on its specific contributions to the main physiological processes in the visual system. Autophagy in the lens As a transparent concentrating structure, the lens achieves cell differentiation through morphological elongation, lens protein expression and organelle programmed clearance [68]. Research has confirmed that autophagy plays a critical role in final tissue organization, homeostatic regulation, and optical transparency [69]. Research on the developing chick embryo lens demonstrates that autophagy drives the formation of the organelle-free zone (OFZ) by degrading cytoplasmic organelles (mitochondria, ER, Golgi), a prerequisite for lens transparency during fiber cell differentiation [70, 71]. Organelle degradation initiates centrally in differentiating fibers and expands peripherally toward the equator. Nuclear elimination involves more complex signaling than non-nuclear organelle clearance, requiring coordinated pathway regulation to complete OFZ formation without inducing cell death [72, 73]. Under conditions of nutritional deficiency or hypoxia, autophagy promotes the clearance of organelles by the activation of PI3K/Akt/mTOR and the inhibition of MAPK/JNK, thereby maintaining homeostasis in lens fibroblasts. The core autophagy markers LC3B-II and Beclin-1 are enriched on vesicles within the central lens area undergoing organelle clearance. Beclin-1 initiates phagophore assembly, while LC3B-II mediates phagophore expansion, cargo recruitment, and autophagosome formation [74]. Critically, both proteins specifically colocalize with degrading organelles within autophagic vesicles. Combined with ultrastructural analyses, these findings demonstrate that autophagy signaling pathways directly mediate nonnuclear organelle degradation for OFZ establishment [71, 72]. The membrane-localized protein PIK3IP1, which structurally resembles the p85 regulatory subunit, interacts with PI3K p110 catalytic domains [71]. Transcriptomic profiling identifies PIK3IP1 as a key transcriptional target in lens fiber cells, where it serves as an endogenous suppressor to promote autophagic clearance during organelle-free zone establishment. During lens development, PIK3IP1 specifically associates with p110α/β/γ subunits in fiber cells concurrently with autophagy-mediated organelle removal [72]. This provides the first evidence that PIK3IP1 spatiotemporally regulates organelle elimination via PI3K-autophagy signaling during OFZ establishment. While animal models provide a vital platform for mechanistic investigations, species-specific differences mean that validation of these mechanisms in human tissues remains scarce. The lens occupies a hypoxic niche due to prenatal regression of its vascular supply, resulting in oxygen dependence on diffusion from the aqueous humor. An anteroposterior oxygen gradient (≈ 2% at surface, declining by ~ 20% toward the core) establishes hypoxia in the organelle-eliminating transition zone [75, 76]. Functional validation confirms HIF1α governs this process: hypoxia (1% O₂) accelerated organelle degradation versus normoxia (21% O₂), while HIF1α stabilization (DMOG) upregulated BNIP3L and enhanced clearanc. On the contrary, inhibiting HIF1α (through Chetomin) can inhibit BNIP3L and hinder its degradation. ChIP-qPCR experiments further confirm that HIF1α can directly bind to the promoter region of BNIP3L [77]. From a pathological point of view, the ATG5 defective lens can remain transparent for the first 5 months after birth, but it begins to become cloudy between 6 and 9 months, and cataracts form at 21 months [78]. Conversely, lens-specific Vps34 deletion causes congenital cataracts and microphthalmia. Normal lens development in ATG5 deficiency versus severe developmental defects in Vps34 knockouts likely reflects the pleiotropic role of Vps34 in endocytic trafficking [69]. Vps34 not only regulates the maturation and selection of endophages and the fusion of late endophages and lysosomes, but also indirectly affects key signaling pathways such as mTOR and EGFR through endophage. Autophagy in the retina As a highly metabolically active neural tissue, the retina requires a large amount of biosynthesis and energy resources to maintain synaptic transmission, visual signal transduction and neuronal membrane potential regulation (Fig. 3). Component autophagy activity plays an indispensable role in visual function; once the activity is damaged, it will disrupt the generation of metabolic intermediates and weaken the retina’s ability to adapt to external stress [79]. Fig. 3The structure of the retina. The retina is a multilayered structure comprising the retinal pigment epithelium (RPE), which closely apposes photoreceptor outer segments. Photoreceptor cell bodies form the outer nuclear layer (ONL) and synapse with bipolar/horizontal cells in the outer plexiform layer. The inner nuclear layer (INL) contains bipolar, horizontal, and amacrine cells, which connect to retinal ganglion cells (RGCs) in the inner plexiform layer. RGC axons constitute the optic nerve, transmitting visual signals to the brain. Autophagy crucially supports RGC survival, optic nerve development, and homeostasis The structure of the retina. The retina is a multilayered structure comprising the retinal pigment epithelium (RPE), which closely apposes photoreceptor outer segments. Photoreceptor cell bodies form the outer nuclear layer (ONL) and synapse with bipolar/horizontal cells in the outer plexiform layer. The inner nuclear layer (INL) contains bipolar, horizontal, and amacrine cells, which connect to retinal ganglion cells (RGCs) in the inner plexiform layer. RGC axons constitute the optic nerve, transmitting visual signals to the brain. Autophagy crucially supports RGC survival, optic nerve development, and homeostasis Zebrafish lacking Ambra1 showed eye deformities, ear hypoplasia and fatal developmental defects [80]. Meanwhile, ATG4B-deficient mice showed partial impairment of autophagy function in multiple tissues, accompanied by the accumulation of ubiquitinized proteins and an increase in p62 levels [81]. Nevertheless, ATG4B deficiency results in preserved retinal electrophysiological responses and unchanged RGC populations in mature mice, indicating that partial autophagy impairment does not substantially affect visual function or ganglion cell development under basal conditions. However, compared with wild-type mice, ATG4B-deficient animals demonstrate greater vulnerability to optic nerve injury and exhibit enhanced neuronal loss following axotomy [82]. Under physiological illumination, retinal architecture and photoreceptor organization remain comparable between Beclin-1 heterozygous and wild-type mice. Following brief high-intensity illumination, however, mutant animals develop extensive retinal pathology, particularly in the superior quadrant, featuring shortened photoreceptor outer segments and outer nuclear layer thinning. Accompanying retinal pigment epithelium (RPE) abnormalities include changes in nuclear morphology, cytoplasmic vacuolization, and mitochondrial swelling accompanied by structural destruction [83]. These experimental results show that under physiological conditions, moderate autophagy damage is compatible with the maintenance of retinal structure and function. However, when confronted with various stress conditions, such as axonal injury in retinal ganglion cells or intense light exposure, which can damage photoreceptors and RPE cells, increased autophagic activity becomes indispensable. This up-regulating effect can not only prevent the accumulation of damaged macromolecules and organelles, but also provide the necessary substrates for energy generation and biosynthesis [83]. AMPK serves essential functions in preserving neuronal viability, particularly in the metabolically active retina. Drosophila studies demonstrate that mutations in the AMPK β-subunit trigger a unique neurodegenerative phenotype featuring vesicle accumulation and photoreceptor loss. This pathological change comes from light-induced metabolic depletion in AMPK-deficient cells, not just the genetic defect itself. Observed cell death requires autophagy activation, and autophagy activation will accelerate neurodegeneration [84]. To investigate developmental cell death within the retina, Linden et al. employed ex vivo preparations derived from postnatal rat retinas [85]. Analysis of the first pathway revealed that individual or combined inhibition of autophagy and caspases resulted in equivalent levels of cytotoxicity, indicating that these processes are coordinated within a common signaling cascade. The second pathway involved caspase-3-mediated mechanisms that execute TUNEL-negative cell death. Complete protection required bongkrekic acid, whereas the effects of caspase inhibitors or 3-MA were only partially rescued, suggesting that the mitochondrial permeability transition initiates dual death cascades, with only one exhibiting autophagic features [86]. Progressive deterioration of visual capacity with advancing age primarily results from retinal tissue degradation and neuronal depletion [87]. Human studies demonstrate attenuated photopic and scotopic retinal responses in senescent individuals through electroretinographic recordings. Furthermore, aging correlates with compromised proteostatic mechanisms, exhibiting diminished efficacy in major degradation pathways encompassing macroautophagy, chaperone-mediated autophagy, and ubiquitin-proteasome mechanisms. Sirtuin 1 (SIRT1) is crucial for safeguarding against age-related retinal degeneration, with decreasing expression levels being directly associated with worsening retinal function. Studies have demonstrated that reduced SIRT1 expression (at both the transcriptional and protein levels) in aged rat retinas is significantly associated with a decreased electroretinogram (ERG) b-wave amplitude [88]. Long-term administration (19 months) of the SIRT1 activator resveratrol effectively reversed this process by upregulating SIRT1 expression, suppressing apoptosis, restoring b-wave amplitude, and activating the BDNF‒TrkB neurotrophic signaling axis to sustain retinal cell survival [88]. Crucially, SIRT1 is involved in the compartmentalized dual regulation of autophagy: it suppresses autophagy via nuclear histone deacetylation but promotes autophagy through cytoplasmic protein deacetylation. Under stress conditions, lysosomal membrane permeabilization (LMP) facilitates the cathepsin-mediated cleavage of SIRT1, subsequently inducing stress-induced premature senescence (SIPS) [89]. This process amplifies autophagic responses through p62 dysregulation, whereas inhibiting cathepsins or blocking LMP prevents SIRT1 depletion [90]. Research has demonstrated that ferulic acid (FA) alleviates radiation-triggered inflammatory responses by upregulating Sirt1 while inhibiting the NLRP3 inflammasome complex. This mechanism effectively decreases secretory cytokine levels and decreases the expression of autophagic markers such as LC3-II. Given that ocular tumors, including retinoblastoma, often lead to retinal damage postradiotherapy, this work elucidates the radioprotective role of FA via cross-talk between autophagy and inflammation, thereby offering potential therapeutic avenues [91]. Age-related retinal degeneration is driven by intertwined pathological mechanisms. Aberrant UPR via PERK activation (elevated pEIF2α, ATF4, GADD34) and ATF6 induction with rhodopsin downregulation; oxidative stress due to weakened NRF2/HO-1 defenses; and inflammation from reduced anti-inflammatory factors (IL-10, IL-13) and increased pro-inflammatory RANTES [92]. Understanding these mechanisms will aid in developing therapies to counteract visual dysfunction linked to aging and age-related retinal diseases. Autophagy in the hyaloid vasculature The vitreous body itself is a transparent gelatinous material without blood vessels. It is mainly composed of water (approximately 99%), a small amount of collagen fibers and hyaluronic acid. Positioned between the lens and the retina, this tissue supports ocular architecture, stabilizes intraocular pressure and facilitates light transmission. In contrast, the hyaloid vasculature represents a transient circulatory network that supplies the developing eye. Complete regression of this system occurs upon maturation of the retinal vasculature [93]. This developmentally programmed involution initiates between postnatal days 4 and 12 and culminates by day 16. The regression process features elevated levels of activated caspase-3 and LC3, demonstrating the concurrent activation of apoptotic and autophagic pathways [94]. Cobalt chloride-induced (CoCl₂) chemical hypoxia initiates temporally regulated autophagy in the immature retina, representing a protective mechanism against metabolic stress that reduces oxidative damage and cellular demise [95]. Although CoCl₂ mimics key features of physiological hypoxia by competitively binding Fe²⁺ to inhibit prolyl hydroxylase (PHD) activity and stabilize HIF-1α, it lacks the local metabolic triggers and precise regulatory signals inherent to physiological hypoxia [96]. Consequently, findings derived from the CoCl₂ model cannot fully recapitulate the physiological outcomes of hyaloid vessel regression. Extrapolation of these results to in vivo processes should therefore be approached with caution and, where possible, complemented with genetic manipulation models (e.g., conditional knockout of vascular endothelial growth factor, VEGF) or alternative hypoxia models for validation [97]. Pharmacological suppression using 3-methyladenine disrupts normal hyaloid vasculature regression. Supporting this, Fusco et al. demonstrated active cell death and vascular endothelial cell immaturity in the P21 rabbit retina. Bevacizumab administration, targeting vascular endothelial growth factor, significantly attenuated both programmed cell death and autophagic activity. Meanwhile, VEGF activates Akt, which suppresses the TSC1/2 complex and thereby relieves inhibition of mTORC1, a principal negative regulator of autophagy [98]. VEGF demonstrates significant neurotrophic properties within neural tissues such as the retina [99]. Although anti-VEGF agents such as bevacizumab show established efficacy in controlling aberrant vessel formation in ocular disorders, including age-related macular degeneration and retinopathy of prematurity [100], their clinical implementation must account for the crucial physiological functions of these molecules in neuronal maintenance and survival. VEGF signaling exhibits a context-dependent duality in the eye. Under physiological conditions, VEGF-A is constitutively expressed by retinal neurons and glia (e.g., astrocytes and Müller cells) [101] and activates VEGFR2 to initiate prosurvival pathways (e.g., PI3K-Akt and ERK1/2). This activity can enhance antioxidant ability, inhibit mitochondrial-mediated apoptosis, and support synaptic function. In the state of pathological hypoxia or inflammation (e.g., in DR or ROP), VEGF is markedly upregulated, shifting its role toward a predominantly angiogenic drive [102]. This process will trigger abnormal vascular proliferation, characterized by the absence of pericytes and an overly permeable vascular system, thereby damaging the integrity of the retinal barrier and ultimately leading to tractional retinal detachment. Incomplete regression of vitreous blood vessels will lead to the persistence of fetal vascular structure, which is clinically manifested as persistent proliferative primary vitreous (PHPV) and accompanied by serious ocular complications. It is worth noting that the autophagy inducer rapamycin can promote the regression of developing vitreous blood vessels in a dose-dependent manner by inducing autophagical cell death in vascular endothelial cells. Evidence from this mechanism level shows that activating the autophagy pathway is expected to be a promising treatment strategy for intervention in PHPV [94]. Autophagy in optic nerve development The optic nerve, which is made up of axons from retinal ganglion cells (RGCs), links the retina to various visual brain centers [103]. These axons create optic nerve fibers within the eyeball, gather at the optic nerve head, and form the optic nerve after traversing the lamina cribrosa. Neural precursors, transitional cells between pluripotent stem cells and mature neurons, can differentiate into RGCs during retinal development [104]. In a mouse model with specific knockout of the ATG5 gene in nerve progenitor cells, researchers observed a significant neurodegeneration phenotype, which led to the death of mice at about 3 months of age. Experimental animals demonstrated progressive deposition of ubiquitinated proteins within cortical, hippocampal, and olfactory bulb neurons, with severity escalating over time [105]. Retinal examination in mutant mice revealed substantial pathological alterations, featuring increased accumulation of the selective autophagy receptor p62, deposition of ubiquitinated proteins within the inner plexiform layer, and TUNEL-positive cells in the outer nuclear layer. The thinning of the ONL of the central and peripheral retinal region confirms the degeneration of the photoreceptor; at the same time, the upregulation of GFAP immune reactivity in the ganglion cell layer and IPL suggests the existence of reactive glial proliferation [106]. Based on this concept, the researchers used AAV2-mediated delivery technology to transport substances to RGCs to explore the effect of inhibiting autophagy. Specifically, silencing ATG5 in RGCs markedly increased vulnerability to optic nerve injury, as shown by lower RGC survival in ATG5flox/flox mice receiving AAV2-GFP-Cre than in AAV2-GFP controls [107]. The E50K variant in Optineurin (OPTN), a selective autophagy receptor responsible for eliminating damaged mitochondria and intracellular pathogens, impairs autophagic processes in retinal ganglion cells and contributes to primary open-angle glaucoma pathogenesis [108, 109]. The transgenic mouse model expressing the mutation showed significant retinal thinning, especially the involvement of the peripheral area of RGCs and the retinal nerve fiber layer around the optic disc was most obvious [110]. In addition, the mutation also leads to an increase in the number of cell apoptosis in each layer of the retina [111]. Transgenic animal models expressing the E50K OPTN mutation demonstrate substantial accumulation of autophagy-related proteins, including LC3 and p62, revealing compromised autophagic progression [112]. This pathological disorder promotes the aggregation of TDP-43 protein, thus exacerbating the degeneration of retinal ganglion cells. Drug-induced autophagy using rapamycin can reduce the pathological changes of TDP-43, enhance the survival rate of RGCs, and alleviate retinal damage caused by mutations [113]. ATG5, OPTN and TDP-43 together form an integrated regulatory network to maintain protein homeostasis and support the survival of RGC. The optic nerve shows dynamic autophagy regulation, and its activity fluctuates with changes in neuronal activity and stress state. Future research, if combined with live cell imaging and dynamic flucs determination technology, will help clarify the space-temporal regulation mechanism of autophagy in the health and pathological state of RGC cells. Autophagy in light transduction Vision begins when photons strike retinal photoreceptors, inducing isomerization of visual pigments and initiating a cascade that transduces light into electrical signals for cortical interpretation. Continuous sight depends on recycling 11-cis-retinal through the visual cycle, a precisely orchestrated exchange between photoreceptors and the RPE [114]. RPE cells provide nutrition, absorb scattered light, and use autophagy to remove damaged organelles and falsely folded proteins, so as to maintain the function and optical clarity of photoreceptors. ATG5 is essential for the degradation of photoreceptor outer segments (POS) and visual pigment regeneration in RPE cells via the LAP pathway. In ATG5ΔRPE mice, created by crossing ATG5flox/flox mice with VMD2-Cre transgenic mice, ATG5 and LC3 lipidation levels are diminished in the RPE, and autophagic flux is disrupted following chloroquine treatment. At the age of 3 months, these mice showed normal retinal function through ERG detection [115]. Although the retinal electrophysiological function remained normal in the early stage, with age, these mutant mice gradually developed progressive photoreceptor dysfunction and were not accompanied by the corresponding loss of neurons. This phenotype is different from the ATG7 defective RPE model; in the latter, the visual function and retinoic acid circulation function of juvenile animals are intact [116]. Kim et al. demonstrated that the RPE-specific loss of ATG5 impairs vitamin A recycling, consequently altering visual performance. Collectively, these findings establish a connection between the atypical ATG5-dependent LAP pathway and retinal function [115, 117]. Concomitantly, ATG5-deficient RPE cells exhibit reduced visual pigment levels. Critically, exogenous administration of visual pigments restores photoreceptor function, demonstrating that ATG5 is indispensable for visual pigment regeneration within the visual cycle. Genetic ablation of FIP200 in the retinal pigment epithelium triggers progressive tissue degeneration in 4-month-old models, characterized by immune cell recruitment, microglial reactivity, lipofuscin deposition, as evidenced by autofluorescent oxidized biomolecules, and progressive outer nuclear layer attenuation [118]. In typical autophagy pathways, these genes play a role in the order from upstream to downstream, forming a precisely regulated casde reaction system. Since FIP200 can completely block the entire autophagy pathway, its absence will lead to the most serious phenotype. Although theoretically speaking, the phenotype caused by ATG7 deletion should be more serious than ATG5 deletion, due to the existence of a partial compensation mechanism, the actual consequences of ATG7 deletion are relatively mild; in contrast, in the model of ATG5 deletion, the phenotype presented is more significant. This difference on the surface is attributed to the functional specificity of each gene. LC3-associated phagocytosis (LAP) represents an alternative degradative mechanism resembling autophagy that functions without the involvement of the ULK1 complex and partially circumvents conventional autophagy initiation, demonstrating minimal dependence on ATG7 [119]. This process nevertheless strictly requires ATG5-facilitated LC3 conjugation to lipid membranes. Its distinctive nature originates from the use of ATG5 to create a specialized, diurnally controlled clearance system optimized for handling the substantial daily influx of phagocytosed photoreceptive components in the retinal pigment epithelium [119], thereby sustaining visual function. Therefore, once the AT G5 function is impaired, it will disrupt the classic autophagy pathway and the LAP pathway at the same time, leading to more serious and comprehensive phenotype consequences. The circadian rhythm regulation of autophagy in the eye is crucial to maintaining the function of phototransfer. Experimental studies in Xenopus demonstrate that light exposure promotes autophagic vesicle generation in photoreceptor cells. This response is eliminated through CRISPR-mediated knockout of GNAT1 or RPE65 genes, confirming that both visual signal transduction and chromophore regeneration pathways regulate light-dependent autophagy activation [120]. Subsequent studies reveal that autophagosome quantity fluctuations correspond to the diurnal translocation patterns of transducin and arrestin between photoreceptor segments during light-dark transitions. Moreover, neither light- nor dark-induced peaks in LC3-II levels are observed in transducin-deficient (Gnat1− /−) or arrestin-deficient (Sag− /−) mice, respectively [121]. Within the retinal pigment epithelium, LC3-associated phagocytosis (LAP) mediates the engulfment and renewal of visual chromophores [115]. This mechanism simultaneously initiates the generation of characteristic bilaminar autophagic structures, which display a dual-phase rhythm synchronized with illumination cycles. Importantly, these regulatory processes differ fundamentally from those observed in photoreceptors [121]. This kind of time-organized mechanism limits the energy-consuming autophagy process to a specific period of time, thus avoiding metabolic conflicts with other ATP-dependent functions such as visual signal transduction and neuronal communication. Moreover, the timely clearance of oxidative waste products prevents toxic accumulation, while large-scale degradation and recycling during relative retinal quiescence supply essential materials (e.g., vitamin A) for daytime photoreceptor outer segment renewal and visual pigment regeneration [122]. Autophagy in the ocular immune environment Maintaining immune balance within the eye is essential for safeguarding vision and preventing pathological injury (Fig. 4). Autophagy plays multifaceted roles in regulating ocular immune responses. Fig. 4Overview of the ocular immune landscape in the healthy human eye. (a) Cornea: Dendritic cells and T lymphocytes reside predominantly in the basal epithelial layer, while stromal immune cells consist mainly of macrophages. (b) Bulbar Conjunctiva: This mucosa-associated lymphoid tissue contains diverse immune cells, including intraepithelial myeloid cells and stromal T and B lymphocytes. (c) Choroid: As a highly vascularized loose connective tissue, it harbors resident dendritic cells, macrophages, and mast cells. (d) Retina: In this neural tissue, resident microglia are predominantly located in the synaptic layers and ganglion cell layer. Autophagy plays an essential role in the ocular immune system by maintaining physical barrier integrity, precisely regulating immune cell functions, and facilitating the clearance of harmful substances while suppressing inflammatory pathways. DC, Dendritic Cell; T cells, T Lymphocytes; Mast cell, mast Cell; Macrophage, macrophage; ILM, inner limiting membrane; NFL, nerve fiber layer; GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer Overview of the ocular immune landscape in the healthy human eye. (a) Cornea: Dendritic cells and T lymphocytes reside predominantly in the basal epithelial layer, while stromal immune cells consist mainly of macrophages. (b) Bulbar Conjunctiva: This mucosa-associated lymphoid tissue contains diverse immune cells, including intraepithelial myeloid cells and stromal T and B lymphocytes. (c) Choroid: As a highly vascularized loose connective tissue, it harbors resident dendritic cells, macrophages, and mast cells. (d) Retina: In this neural tissue, resident microglia are predominantly located in the synaptic layers and ganglion cell layer. Autophagy plays an essential role in the ocular immune system by maintaining physical barrier integrity, precisely regulating immune cell functions, and facilitating the clearance of harmful substances while suppressing inflammatory pathways. DC, Dendritic Cell; T cells, T Lymphocytes; Mast cell, mast Cell; Macrophage, macrophage; ILM, inner limiting membrane; NFL, nerve fiber layer; GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer Corneal epithelium is a crucial defense barrier in the anatomical structure of the eye, which is composed of five to seven layers of squamous epithelial cells. The integrity of its structure depends on closely connected proteins such as Claudin-1 and Occludin, which form a sealed structure between cells; therefore, the effectiveness of this barrier function is crucial to maintaining the transparency and physiological function of the cornea. Using transgenic mice labeled with LC3 expressing green fluorescent protein (GFP) to evaluate the state of autophagy in the cornea, the study found that there were more LC3 positive spots contained in corneal margin epithelial basal cells than central corneal epithelial basal cells, which indicates that corneal margin basal epithelial cells have stronger autophagy activity [123]. Functionally, the corneal margin epithelium is the main reservoir for corneal epithelial renewal, which plays a crucial role in maintaining tissue homeosis. The area promotes wound healing and tissue regeneration by continuously supplying progenitor cells [124]. As the main physical barrier in the immune microenvironment of the eye, the functional integrity of the blood retinal barrier depends on autophagy. By mediating the removal of damaged mitochondria and other organelles, autophagy maintains the homeosis of the internal environment of retinal cells and protects the structure of the blood retinal barrier. In diabetic retinopathy, oxidative stress and inflammatory cascadation reactions induced by hyperglycemia will destroy the barrier function; while enhanced autophagy activation can provide protection against such damage [125]. The BRB expresses several important efflux transporters, including P-glycoprotein, breast cancer resistance protein, and multidrug resistance proteins, which critically regulate drug and metabolite clearance from retinal tissue [126–128]. Importantly, autophagy induction can modulate both expression and functional activity of these transport systems, consequently influencing pharmaceutical distribution within retinal compartments. As principal immune effector cells in the central nervous system, microglia demonstrate remarkable functional adaptability and are capable of transitioning between proinflammatory M1 and anti-inflammatory M2 polarization states. Autophagy serves as an essential modulator of this phenotypic conversion. Experimental data from acute glaucoma models indicate that irisin stimulates integrin αVβ5 signaling and AMPK activation, substantially increasing autophagic flux. This increase in autophagy levels promotes the differentiation of small glial cells into neuroprotective M2 phenotype, thus inhibiting neuroinflammatory reactions and limiting the degeneration of retinal ganglion cells [129]. In retinopathy of prematurity, hypoxia-triggered autophagy fosters AIM2 inflammasome assembly, steering microglia toward an M1 phenotype through ASC/CASP1/IL-1β signalling and thereby intensifying pathological retinal neovascularisation [130]. These findings reveal that autophagy regulates the phenotype of microglia bidirectionally by degrading inflammation-related proteins (such as AIM2). Different parts of the eye, including the anterior chamber, retina, and cornea, possess unique immune-privileged characteristics. Micro-environmental factors in these regions can significantly affect autophagy activity through a variety of signal axes, thus directly reshaping the immune pattern in the eye environment [131]. Autophagic dysfunction leads to accumulation of damaged organelles and harmful substances, thereby triggering pro-inflammatory signaling pathways that exacerbate neuroinflammation. Experimental studies demonstrate that in DR, artesunate activates the AMPK/SIRT1 axis to enhance autophagic flux. This therapeutic approach suppresses microglia-mediated secretion of inflammatory mediators such as IL-1β and TNF-α while decreasing reactive oxygen species production, thereby mitigating compromise of blood-retinal barrier integrity [132]. Conversely, in retinitis pigmentosa (RP), oxidative stress suppresses autophagy-dependent protein degradation, resulting in toxic protein aggregates (e.g., ubiquitinated inclusions) within microglia. These aggregates further activate the P2 × 7 receptor, amplifying inflammatory responses and accelerating photoreceptor degeneration [133]. Based on the above findings, in the retinopathy (ROP) model of premature infants, pathological neovascularization mediated by small glial cells was significantly inhibited by targeted knockout of AIM2 or IL-1β inhibitors; thus confirming that the regulation of the neuroimmune pathway through autophagy pathway has potential application value in the treatment of retinal disease [130]. In addition to the regulation of small glial cells, autophagy also plays a crucial role in regulating the function of dendritic cells (DC) by regulating antigen presentation and T cell immunity. Hereditary knockout of the core autophagy gene Atg5 will damage the maturity of DC and weaken its antigen presentation ability, thereby inhibiting the activation of T cells [134]. Autophagy modulates the abundance of co-stimulatory ligands such as CD80/CD86 and of MHC molecules on dendritic-cell membranes, thereby calibrating the efficiency of T-cell activation. For instance, autophagy-deficient DCs exhibit upregulated CD36 surface receptors, which enhance phagocytosis of apoptotic tumor cells while paradoxically suppressing T-cell responses [134]. Dendritic cell autophagy sustains immunological tolerance and Treg stability, governing cellular maturation, antigen processing, and immunomodulation, thereby offering novel therapeutic avenues for ocular pathologies [135]. Autophagy in ocular diseases Autophagy has context-dependent functions in ophthalmic pathology. It maintains retinal homeostasis through the clearance of oxidative damage components and metabolic regulation. In a dysfunctional state, abnormal autophagic flux will weaken lysosomal function, resulting in lipofucin deposition, protein aggregation and activation of inflammatory pathways. These mechanisms promote the progression of geographic atrophy, glaucomatous neurodegeneration, and diabetic retinal vasculopathy, fundamentally driven by a self-perpetuating cycle of lysosomal insufficiency and pathogenic substrate accumulation. Autophagy in age-related macular degeneration AMD, a bilateral retinal disorder affecting central vision, represents a leading cause of irreversible visual impairment in elderly populations. Epidemiological data indicate 25.3% of adults over 60 develop early-stage AMD, with 2.4% progressing to advanced disease. These rates are projected to rise with demographic aging [136]. Clinically, AMD manifests as two primary subtypes. The dry form comprises about 75% of diagnoses and presents with drusen accumulation and central retinal atrophy. In contrast, wet AMD involves pathological choroidal neovascularization driven by upregulated VEGF signaling, leading to subretinal fluid accumulation and hemorrhage [137]. The pathogenesis of AMD involves multiple determinants, which is the result of the joint action of genetic susceptibility and environmental factors. Old age is the main susceptibility factor, while smoking, high-fat diet and high blood pressure are other risk factors. The autophagy process is essentially related to the pathological progression of AMD (Fig. 5). Retinal pigment epithelium deterioration serves as an initial pathological indicator in AMD. These specialized cells perform phagocytic clearance and metabolic processing of discarded photoreceptor components during visual cycling, thereby preserving photoreceptor equilibrium. Previous studies have confirmed that a deficiency in ATG5 in RPE cells impairs their ability to degrade POSs, leading to metabolic dysregulation [138, 139]. Fig. 5The Role of Autophagy in Ocular Diseases. This schematic illustrates the complex involvement of autophagy in the pathogenesis of various ocular disorders, including age-related macular degeneration, glaucoma, cataract, diabetic retinopathy, retinal dystrophies, and retinal detachment. It particularly emphasizes how autophagy exerts dual roles in maintaining cellular homeostasis, metabolic reprogramming, immune surveillance, and regulation of cell death pathways, thereby critically influencing disease progression. mTOR, mechanistic target of rapamycin; AMPK, AMP-activated protein kinase; Beclin1, Beclin1; Atg, autophagy-related gene; LC3, microtubule-associated protein 1 A/1B-light chain 3; HIF-1α, hypoxia-inducible factor 1-alpha; ROS, reactive oxygen species; ​​Ca²⁺​, calcium ion; A2E, N-retinylidene-N-retinylethanolamine; HSP70, heat shock protein 70; HDL, high-density lipoprotein; DNA, deoxyribonucleic acid; RPE, retinal pigment epithelium; LMP, lysosomal membrane permeabilization; ER, endoplasmic reticulum; TXNIP, thioredoxin-interacting protein; NF-κB, nuclear factor kappa-light-chain-enhancer of activated B cells; PINK1, PTEN-induced putative kinase 1; SIRT1, sirtuin 1; FOXO, forkhead box O; HMGB1,high-mobility group box 1; ERCC6; excision repair cross-complementation group 6; VCP, valosin-containing protein The Role of Autophagy in Ocular Diseases. This schematic illustrates the complex involvement of autophagy in the pathogenesis of various ocular disorders, including age-related macular degeneration, glaucoma, cataract, diabetic retinopathy, retinal dystrophies, and retinal detachment. It particularly emphasizes how autophagy exerts dual roles in maintaining cellular homeostasis, metabolic reprogramming, immune surveillance, and regulation of cell death pathways, thereby critically influencing disease progression. mTOR, mechanistic target of rapamycin; AMPK, AMP-activated protein kinase; Beclin1, Beclin1; Atg, autophagy-related gene; LC3, microtubule-associated protein 1 A/1B-light chain 3; HIF-1α, hypoxia-inducible factor 1-alpha; ROS, reactive oxygen species; ​​Ca²⁺​, calcium ion; A2E, N-retinylidene-N-retinylethanolamine; HSP70, heat shock protein 70; HDL, high-density lipoprotein; DNA, deoxyribonucleic acid; RPE, retinal pigment epithelium; LMP, lysosomal membrane permeabilization; ER, endoplasmic reticulum; TXNIP, thioredoxin-interacting protein; NF-κB, nuclear factor kappa-light-chain-enhancer of activated B cells; PINK1, PTEN-induced putative kinase 1; SIRT1, sirtuin 1; FOXO, forkhead box O; HMGB1,high-mobility group box 1; ERCC6; excision repair cross-complementation group 6; VCP, valosin-containing protein The engulfment of photoreceptor outer segments represents an energetically demanding mechanism requiring considerable oxygen consumption. Coupled with persistent light exposure generating photo-oxidative damage, this process imposes substantial oxidative challenges on retinal pigment epithelial cells [140]. In the face of metabolic challenges and oxidative damage, changes in environmental risk factors play a two-way regulatory role through autophagy pathways. A brief redox imbalance can initiate a cell-protective autophagy response, and long-term or severe oxidation load will eventually damage autophagy function. Research by Mitter et al. revealed that short-term hydrogen peroxide treatment can enhance autophagy activity; among them, controlled autophagy induction acts as a cell protection mechanism, helping to remove damaged organelles and reduce redox imbalance [141]. This interpretation receives additional validation from ARPE-19 cellular models, where elevated ethanol concentrations stimulate both enhanced autophagy progression and upregulated expression of multiple autophagic pathway components [142, 143]. However, chronic exposure to the same concentration of hydrogen peroxide (administered every 24 h for 14 days) resulted in reduced autophagic flux. In each condition, autophagy operated protectively; blocking it with 3-MA or Beclin-1/ATG7 shRNA heightened oxidative stress, lowered mitochondrial membrane potential and depleted glutathione, effects that rapamycin reversed [141]. Tobacco smoke, largely through its induction of oxidative damage, represents a major environmental determinant for age-related macular degeneration pathogenesis. Cytotoxic concentrations of cigarette smoke constituents suppress proteasomal function while activating integrated cellular defense mechanisms in retinal pigment epithelium. This response involves compensatory enhancement of autophagic activity and elevated expression of multiple cytoprotective and autophagy-associated genes including p62 [144]. Genetic suppression of p62 elevates carbonylated protein accumulation while reducing nuclear factor erythroid 2-related factor 2 (Nrf2) antioxidant signaling [145]. ARPE-19 cultures exposed to tobacco smoke demonstrate increased transcriptional activation of vascular endothelial growth factor and its receptors VEGFR-1 and VEGFR-2. Combined treatment with recombinant VEGF and cigarette smoke components further augments autophagic flux and promotes cellular viability [144]. Ultraviolet exposure constitutes an additional environmental risk factor for age-related macular degeneration. UV-irradiated retinal pigment epithelial cells serve as established in vitro disease models. Following UVB exposure, the efficacy of antioxidant and neuroprotective agents is abolished, while apoptotic processes are suppressed. These cytoprotective compounds also demonstrate reduced effectiveness in cells lacking ATG5 expression [146]. The intense metabolic activity of RPE cells leads to the progressive accumulation of undigested photoreceptor outer segment material within lysosomal compartments with age, forming lipofuscin composed of oxidized protein-lipid complexes [147]. Its primary constituent, N-retinylidene-N-retinylethanolamine (A2E), both impairs lysosomal enzymatic function and resists degradation, resulting in sustained lysosomal dysfunction. This pathological feature, together with impaired autophagic flux, is commonly observed across all AMD-related pathologies, suggesting that impaired autophagy and lipofuscin deposition establish a self-perpetuating vicious cycle that drives disease progression [133, 148]. In animal models, Toops et al. observed abnormal accumulation of A2E and other retinoid metabolic byproducts within RPE cells in Abca4 knockout mice. In primary cultured mouse RPE cells, exogenous treatment with the bis-retinoid A2E recapitulated the phenotype of reduced autophagic activity; conversely, administration of TO901317 to activate cholesterol transport proteins effectively restored autophagic flux [149]. These findings collectively implicate a central pathological triad in which lysosomal impairment, lipofuscin accumulation, and autophagy suppression form a self-perpetuating vicious cycle that drives RPE degeneration [141, 150]. The molecular basis of this cycle lies in the direct damage caused by A2E to lysosomal function. In vitro studies have further confirmed that the fluorescent compound A2E specifically inhibits the activity of the lysosomal proton pump, significantly elevates intra-lysosomal pH, and consequently impairs hydrolytic enzyme function [151]. This effect is further amplified upon co-incubation with peroxidized palmitic acid derivatives, leading to disrupted lysosomal homeostasis and blockade of the autophagy-mediated protein degradation pathway [150]. Furthermore, the accumulation of A2E will also increase the level of ceramides, disrupt the dynamics of the microtubule network, and impair the cells’ ability to remove waste on multiple levels [152]. Investigations employing the human retinal pigment epithelial line ARPE-19 have established that pharmacological suppression of autophagy using 3-methyladenine promotes lipofuscin formation [143], while rapamycin administration reduces its deposition [141]. A concentration-dependent response was documented in MDA-treated ARPE-19 cultures: lower concentrations potentiated autophagic activity and increased expression of autophagy-related proteins, whereas elevated doses impaired autophagic progression [141, 153]. These findings collectively demonstrate that retinal pigment epithelium primarily activates autophagic pathways in response to diverse forms of cellular stress. Mechanistic investigations have revealed that pharmacological inhibition of acid sphingomyelinase (ASM) can reverse A2E-induced impairment of autophagic flux. These findings demonstrate that bis-retinoid components of lipofuscin disrupt RPE autophagic homeostasis, and that targeting ASM restores autophagic activity [149]. However, complete loss of ASM function (as in ASM knockout mouse models) also leads to visual impairment, lipofuscin accumulation, and reduced autophagic flux [154]. Together, these results indicate that precise balance within the retinal cholesterol-ceramide metabolic network is critical. Thus, any intervention that disrupts this equilibrium, whether enhancing or suppressing ASM activity, significantly compromises lysosomal function, ultimately resulting in impaired lipofuscin clearance, dysregulated autophagy, and retinal functional damage. It should be acknowledged that ARPE-19, as a transformed cell line, incompletely recapitulates the progressive autophagic decline observed in human aging and AMD. The field requires models capable of simulating graded autophagy impairment to identify critical therapeutic windows [155]. Drusen formation, a fundamental initial pathological feature in age-related macular degeneration, involves autophagic processes that participate significantly in their development. Research shows that knocking down AMPK or TAK-1 through genetic means to block AMPK signal conduction and the subsequent autophagy induction process will enhance the susceptibility of cells to death. Further supporting this concept, rotenone-exposed ARPE-19 cells exhibit diminished cathepsin D function, demonstrating lysosomal impairment [156]. After mitochondrial damage in ARPE-19 cell culture, the upregulation of the expression of exocrine and lysosomal markers (CD63 and Lamp2) suggests that when the lysosomal degradation function is impaired, exocrine biosynthesis may be used as an alternative way to remove toxic protein aggregates. This mechanism is expected to alleviate cytoplasmic stress. Corresponding observations in aged murine RPE cells revealed autophagic markers such as ATG12 immunoreactivity alongside elevated exosomal indicators, suggesting that extracellular vesicle-mediated expulsion of engulfed components may participate in drusen deposition [156]. Diverse in vivo and in vitro models, ranging from genetic knockout to phenotypic alterations, have consistently demonstrated that perturbation of critical autophagic nodes modulates fundamental pathological processes underlying AMD. Research utilizing C57 mice with SOD2 knockout to model redox imbalance revealed enhanced autophagic activity during early developmental phases (1-month), succeeded by progressive downregulation of ATG7, ATG9, and LC3 commencing at three months. Parallel investigations of human donor specimens with advanced AMD exhibited diminished autophagy-related protein expression at comparable stages [141]. Complementary studies established that Nrf2-deficient animals develop pathological features resembling AMD, including deposition of drusen-like material between 8 and 11 months, compromised retinal physiology, accumulated lipofuscin and undegraded photoreceptor components, elevated oxidative stress indicators, autophagic dysregulation, and mitochondrial impairment [157]. The CreVMD2 murine system facilitates tissue-specific genetic ablation in retinal pigment epithelium through a BEST1 promoter directing tissue-restricted transgene expression [158]. Conditional elimination of the essential autophagy initiation factor RB1CC1/FIP200 in RPE triggered primary epithelial dystrophy, structural anomalies, and buildup of autofluorescent lipofuscin with oxidized proteins. These pathological alterations included cytoplasmic inclusion bodies, microglial recruitment, and age-related Bruch’s membrane deterioration, ultimately progressing to secondary photoreceptor degeneration and visual impairment [118]. As an inhibitory factor upstream of the mTORC1 signaling pathway, TSC1 defects can trigger pathological signs of wet and dry AMD. In the mouse model and the primary human retinal pigment epithelium (RPE) cell culture system, the gene knockout of TSC1 will lead to a series of characteristic lesions, including abnormal fundus spontaneous fluorescence signals, progressive choroidal retinal atrophy, and final retinal degeneration [159, 160]. Clinical analyses reveal higher p62 accumulation in the macula of AMD patients versus controls, implying compromised RPE autophagy and reinforcing the concept that flux blockade contributes to AMD development [161]. Furthermore, investigations document heightened malondialdehyde concentrations in both circulating blood and RPE specimens from affected individuals [153]. These oxidative stress markers correlate with enhanced LC3-II conversion and diminished p62 degradation respectively, supporting a pathogenic model where adaptive autophagic induction during initial phases eventually progresses to functional exhaustion in advanced stages. Recent clinical studies have shown that local chronic inflammation driven by metabolic disorders of RPE cells is the key to the progression of AMD, and impaired autophagic flux plays a core driving role by accelerating the signal transduction of the NLRP3 inflammasome [54, 162]. Rotenone-mediated RPE injury triggers inflammasome assembly and subsequent pro-inflammatory cytokine secretion in macrophages, additionally stimulating pathological blood vessel formation. Intravitreal administration of wortmannin, an autophagy suppressor, promotes RPE degeneration and recruits inflammation-activated macrophages. These findings indicate that autophagic dysfunction in age-related macular degeneration potentially amplifies pro-inflammatory pathway activation through inflammasome signaling and angiogenic factor release, establishing a self-perpetuating cycle connecting impaired autophagy with chronic inflammation [163]. Both proteasomal dysfunction and autophagic disruption enhance IL-1β secretion in human RPE cultures, subsequently promoting NLRP3 inflammasome assembly [164]. While inflammasome activation contributes to AMD progression, its potential role in causing autophagic dysregulation requires further investigation [165, 166]. Autophagy in glaucoma Glaucoma is a slowly progressive neurodegenerative disease characterized by irreversible loss of retinal ganglion cells, resulting in visual field defects and blindness. Its insidious course often delays detection until advanced stages, and it is the world’s second leading cause of blindness [167]. This multifactorial disorder results from the interplay of multiple molecular pathways, with elevated intraocular pressure (IOP) and aging as primary risk factors [168]. Accumulating experimental evidence indicates that autophagic activity progressively declines in both the brain and retina as a function of aging [169]. This age-related decline contributes to various degenerative conditions such as glaucoma (Fig. 5) [120]. Investigations using AMBRA1-deficient murine models displaying impaired autophagic function following optic nerve injury demonstrated progressively exacerbated retinal ganglion cell vulnerability with advancing age [170, 171]. It is worth noting that older animals (13 months old) showed higher neuronal susceptibility than younger individuals (3 months old) [170]. AMBRA1 (autophagy and beclin-1 regulator 1) facilitates autophagosome formation and serves as a crucial mitochondrial quality control receptor [171], functioning through both PINK1/PARKIN-mediated and alternative pathways [172, 173]. Notably, compared with age-matched wild-type controls, young wild-type and AMBRA1+/gt mice exhibited comparable RGC survival, whereas aged AMBRA1+/gt animals exhibited significantly greater retinal ganglion cell loss after optic nerve crush [170]. This enhanced vulnerability may be attributable to a diminished oxidative stress response postinjury, coupled with impaired upregulation of Bnip3 and compromised Bnip3l-mediated autophagic activity. Early research on glaucoma mainly focuses on the molecular pathway that leads to the degeneration of retinal ganglion cells, aiming to establish a neuroprotection intervention strategy that can play an independent role or in conjunction with intraocular pressure regulation. A variety of pathological mechanisms can lead to retinal ganglion cell degeneration, including lack of neurotrophic support, redox imbalance and exciting neurotransmitter toxicity [174, 175]. Experimental evidence from animal models of glaucoma shows that exogenous administration of neurotrophic factors or redox stabilizers can effectively reduce the loss of neurons [175]. Additional insults, including chronic hypoxia and disruption of axonal transport, impose disproportionate metabolic demands on RGCs. Due to the extremely long axon projection of these cells and the lack of myelin sheath, they show high dependence on strong mitochondrial function, which significantly increases their susceptibility to oxidative damage [176, 177]. Genome-wide association data indicate that several glaucoma-linked loci, including Optn and Tbk1, encode proteins that directly regulate autophagy and mitochondrial homeostasis [178, 179]. Autophagy has multifaceted and condition-specific functions in glaucomatous pathology. When suffering from cellular stress, retinal ganglion cells will activate this self-degrading pathway as a compensatory mechanism to resist environmental challenges and maintain intracellular homeosis. A large amount of evidence shows that the activation of autophagy has a clear protective effect after axon damage to RGCs. Studies using optic nerve transection (ONT) models have demonstrated that rapamycin treatment increases RGC survival by 40% at 10 days post-axotomy [107]. Research on ONT models further revealed that Atg4b⁻/⁻ mice exhibited 28% greater RGC loss compared to wild-type controls, while RGC-specific deletion of Atg5 similarly resulted in a 33% reduction in RGC survival after ONT [107]. These findings from multiple model systems collectively indicate that promoting autophagy activation enhances RGC protection and significantly improves cell survival rates. This effect is likely mediated through mechanisms such as alleviation of oxidative stress and clearance of damaged organelles—e.g., via mitophagy. In a study examining the acute effects of optic nerve crush in rats, axons responded with a rapid Ca²⁺ influx and focal swelling within 30 min [180]. Ultrastructural analysis revealed disorganized microtubules and neurofilaments, dysmorphic mitochondria, and numerous LC3-positive autophagosomes along the injured fibers. Because Ca²⁺ dyshomeostasis accompanies many neurologic disorders and because traumatic axotomy is known to open plasmalemmal Ca²⁺ channels, these data suggest that extracellular Ca²⁺ is an early trigger of axonal autophagy [181]. To test the link functionally, animals were given an intravitreal injection of 3-methyladenine (3-MA), an autophagy blocker, 48 h before the crush. Drug treatment significantly slowed axonal fragmentation at the 6-h time-point, indicating that autophagy accelerates acute axonal disintegration [180]. Consistent with a Ca²⁺-dependent mechanism, co-administration of Ca²⁺-channel blockers reduced both LC3-II and p62 immunosignals, supporting the notion that intra-axonal Ca²⁺ levels mediate the rapid induction of autophagy following traumatic axonal injury. In contrast, Kitaoka et al. reported that rapamycin treatment produced significant axon-sparing effects in a laser-induced ocular hypertension rat model, whereas pharmacological suppression of autophagy with 3-methyladenine accelerated axonal deterioration [182]. However, a separate investigation in which episcleral venous occlusion was used to establish glaucoma revealed improved neuronal survival following 3-MA-mediated autophagy inhibition [183]. These divergent outcomes may reflect distinct intervention timelines in the experimental protocols. The neuroprotective potential of rapamycin-induced autophagy was validated in a murine retinal ischemia‒reperfusion injury system [184]. These contradictory findings highlight that autophagic outcomes are highly sensitive to the timing of intervention and the dose of modulators. Given the distinct microenvironments of neuronal compartments, further studies are needed to determine whether autophagy is differentially regulated between the somata and axons of retinal ganglion cells. To clarify this paradoxical observation, investigators detected increased oxidative damage in experimental glaucoma models and proposed that enhanced autophagic activity may alleviate redox imbalance [185]. Consistent with this hypothesis, genetic ablation of autophagy pathways results in reactive oxygen species accumulation, potentially through defective removal of oxidation-damaged cellular constituents [185, 186]. However, severe oxidative stress preferentially compromises lysosomal membrane integrity, triggering cathepsin translocation to the cytosol [187]. Beyond the cytotoxic effects of mislocalized proteolytic enzymes, compromised lysosomes become incapable of autophagosome fusion, ultimately reducing functional autophagic degradation despite elevated autophagic vesicle quantity. Increased intracellular calcium concentrations produce analogous pathological effects: following axonal damage, calcium influx stimulates the expression of calpain, which mediates the proteolytic processing of core autophagic components, including ATG5 and Beclin-1, consequently suppressing autophagic progression. This process additionally induces lysosomal membrane disruption, promoting cathepsin release and ultimately triggering cellular death [188]. Importantly, pharmacological blockade of calpain activation, cathepsin activity, or lysosomal function can prevent these degenerative cascades and preserve RGC viability [189]. Ca²⁺ overload or oxidative stress can rupture lysosomal membranes, release cathepsins and halt autophagic flux. Under these conditions, autophagosomes accumulate, but their cargo cannot be degraded, resulting in ‘maladaptive autophagy’ that is toxic rather than protective. This scenario explains at least to some extent why autophagy is regulated to produce completely different results between different experimental models, time points and neuronal compartments (e.g., soma versus axon) [190]. In addition, glaucoma is a disease involving a variety of cell types, and autophagy also plays an important role in addition to retinal ganglion cells. In experimental glaucoma models, microglia demonstrate adaptive responses to pathological microenvironments, potentially providing neuroprotective functions through clearance of neuronal debris and secretion of trophic/survival-promoting factors. However, continuous pathological stimulation will induce the transformation of the phenotype to a harmful activated state, which is characterized by the secretion of cytotoxic mediators and the removal of neuronal components through phagocytic action [191]. Optic nerve astrocytes significantly participate in glaucomatous pathology. Studies have indicated that astrocytes in ocular hypertensive rats exhibit inflammatory activation markers, including TNF-α/NF-κB pathway stimulation and increased ATG7 expression [192]. Further research reveals that these cells engulf and remove axon mitochondria through a special process called ‘transmitophagy’ [193]. Notably, astrocytes normally express phagocytic regulators [191] such as Mac-2, whose expression is upregulated in glaucoma [177], although the precise involvement of autophagic mechanisms in these cells requires further investigation. Trabecular meshwork (TM) cells form a porous, mesh-like network of channels responsible for controlling the outflow of aqueous humor, thereby regulating intraocular pressure (IOP). Autophagy is crucial to this steady-state regulation. Under physiological conditions, TM cells are continuously subjected to mechanical stretch exerted by flowing aqueous humor, oxidative free radicals generated by metabolic activities, and phagocytic stress caused by the continuous removal of pigments, cell fragments and protein aggregates. In vitro, porcine TM cells exposed to hyperoxia and oxidative stress exhibit a marked cessation of autophagic flux, accompanied by lysosomal deacidification and impaired protease activity [194, 195]. Moreover, studies in human tissue have confirmed that compared with those from healthy donors, trabecular meshwork (TM) cells from glaucoma patients exhibit hallmark features of impaired autophagy, including significantly reduced LC3-II levels, diminished lysosomal proteolytic function, and markedly elevated β-galactosidase activity and lipofuscin accumulation [195]. However, current research on glaucoma models faces certain limitations, as many of the animal models used fail to fully recapitulate the human disease conditions [196]. The DBA/2J murine model progressively develops asymmetrical ocular hypertension [197], whereas acute glaucoma paradigms utilize impeded aqueous drainage, hypertonic solutions or microbeads, venous occlusion techniques, or intravitreal NMDA administration to rapidly induce intraocular pressure elevation and retinal ganglion cell injury [196, 198]. Consequently, data analysis should carefully consider the pathophysiological relevance and inherent limitations of each methodology. Autophagy in diabetic retinopathy DR, the leading microvascular ocular complication of diabetes, results from chronic hyperglycemia that triggers retinal inflammation, endoplasmic reticulum (ER) stress, mitochondrial impairment and oxidative injury at early disease stages [125, 199]. Autophagy’s involvement in diabetic pathology is well-documented. Type 2 diabetes pathogenesis involves insulin resistance and functional impairment of pancreatic β-cells. Conditional knockout of ATG7 in β-cells results in diminished pancreatic β-cell population, mitochondrial dysfunction, and elevated reactive oxygen species generation, promoting insulin resistance and secretion defects [200, 201]. Additionally, autophagy mediates cellular protection during hyperglycemic conditions by responding to endoplasmic reticulum stress and eliminating ubiquitinated protein clusters from β-cells [202, 203]. The role of autophagy in diabetic retinopathy varies significantly depending on the stage of pathological progression and the cellular environment (Fig. 5). In the early stages, this process exerts a cytoprotective effect. Within diabetic retinal tissue, the unfolded protein response is activated, thereby promoting inflammatory processes, programmed cell death, and neovascularization [204, 205]. As a basic adaptation mechanism to endoplasmic reticular stress, autophagy significantly affects the pathogenesis of DR pathogenesis. Stimulation of UPR signaling pathways, particularly the IRE1α/XBP1 axis, mitigates stress damage and affects diabetic retinal pathology. Within this framework, XBP1 potentially regulates inflammatory mechanisms to inhibit retinal inflammation [205, 206]. while simultaneously controlling vascular proliferation and enhancing endothelial cell growth [207]. Additionally, it can initiate autophagy through acetylation-dependent modification of ATG9A [208], representing a promising treatment approach. Recent investigations have demonstrated that autophagic activity is stimulated in ARPE-19 cultures under elevated glucose conditions; pharmacological inhibition using 3-methyladenine causes cellular survival to significantly decrease, the accumulation of reactive oxygen species, and the secretion of proinflammatory interleukin-1β in conjunction with NLRP3 inflammasome activation [209]. This self-degradative process consequently provides cellular protection through the clearance of oxidative damage components and the suppression of inflammatory activation, potentially representing a valuable therapeutic approach for inhibiting the initiation and development of diabetic retinopathy. Preclinical investigations using diabetic rodent models demonstrate elevated thioredoxin-interacting protein expression in retinal tissue. Via NF-κB it amplifies inflammation (IL-1β) and increases ROS, leading to protein misfolding and organelle damage [210]. Because TXNIP simultaneously engages pro-survival pathways, overall cell viability is preserved. It drives mitophagy via PINK1/PARKIN to clear ROS-injured mitochondria, co-localises with LC3, exports HMGB1 to the cytosol, and partners with Beclin-1 to trigger autophagy [211]. Under sustained and severe stress, however, autophagy switches from protective to detrimental. Pericyte loss is an early hallmark of DR. When blood-retinal barrier integrity becomes impaired, modified low-density lipoprotein infiltrates retinal tissue, inducing both oxidative and endoplasmic reticulum stress. This process enhances autophagic activity and promotes programmed cell death in retinal pigment epithelium and pericyte populations [212]. The effect of modified LDL on pericytes demonstrates concentration-dependent biphasic characteristics​​. Under ​​low-dose conditions​​, pericytes can induce ​​cytoprotective autophagy​​, thereby maintaining cellular viability. In contrast, under ​​high-dose exposure​, this protective autophagic response ​​is no longer observed​​, and conversely, ​​the inhibition of autophagy rescues pericytes from cell death​ [212, 213]. Recent work has shown that retinal ischemia disrupts interpericyte communication, altering blood flow and neurovascular coupling. Pericytes are interconnected by tunneling nanotube-like structures that allow the transfer of vesicles and even mitochondria [214], offering a novel mechanism for pericyte dysfunction, although the impact of this process on autophagy in DR remains to be defined. Autophagy in cataracts Cataracts are a disorder in which the lens loses transparency or changes color, scattering or blocking light and thereby reducing visual acuity. The core pathology involves lens-protein denaturation, aggregation or precipitation, which converts the clear lens into an opaque structure [215]. The principal risk factors include age-related protein alterations (the most common cause), genetically determined congenital forms, trauma, secondary changes linked to diabetes or corticosteroid use, and radiation exposure [4]. Under the background of congenital cataracts, the structure of the lens is composed of single-layer cubic pre-epithelial cells and their terminally differentiated fibroblasts. In the process of cell maturation, membrane surrounding structures, including mitochondria and endoplasmic reticulum, will undergo a removal process, thus forming an area that lacks organelles. Damage to this clearance mechanism will lead to abnormal development and loss of optical clarity [216, 217]. This process is crucial to maintain the transparency of the lens (Fig. 5). Recent studies have shown that the genetic mutations of FYCO1, Pik3c3, CHMP4B, EPG5, ERCC6 and TDRD7 genes can damage the autophagy process, resulting in hereditary lens opacity. Among them, FYCO1 mutation specifically causes autosomal recessive hereditary congenital cataract. As a key junction protein, FYCO1 interacts with LC3 and RAB7 at the same time to promote microtubule-based autophagy transport, thus guiding the vesicles to move to the plus end of the microtubule for lysosomal degradation [218]. Consequently, FYCO1 mutations disrupt autophagosome trafficking. Although the transcription and translation of autophagy-related components are upregulated in lens cells, the degradation pathway remains functionally obstructed. CHMP4B, one of three mammalian orthologs of the human VPS32 gene, is implicated in autosomal dominant congenital cataracts through mutations that disrupt the degradative phase of autophagic flux [219]. As an integral subunit of the ESCRT-III machinery, CHMP4B contributes to maintenance of endolysosomal membrane [220, 221], endoplasmic reticulum-selective autophagy, and phagophore sealing during mitochondrial clearance [222]. Genetic studies demonstrate that complete knockout of murine Chmp4b causes embryonic mortality, while its deficiency results in autophagic vesicle accumulation in Drosophila visual systems and mammalian cortical neurons. Heterologous expression experiments have determined that a CHMP4B mutant (p.D129V) associated with cataract disrupts endogenous transport [223] and impairs chromatin interaction, suggesting that CHMP4B participates in the autophagic lysosomal degradation process of micronucleated and extranuclear genetic material [219]. EPG5, the human ortholog of epg-5, encodes an essential factor controlling autophagosome maturation and autolysosome formation. Pathogenic variants in EPG5 are established as causative for Vici syndrome, a multisystem hereditary condition that may manifest with congenital cataracts [224]. ERCC6, a critical component of the nucleotide excision repair (NER) DNA repair mechanism, interacts with the lysosomal membrane protein valosin-containing protein (VCP). This interaction regulates autophagosome-lysosome tethering and fusion. Once the process is disturbed, the autophagy degradation function in the lens epithelial cells will be impaired [225]. Studies have indicated that TDRD7 can directly interact with and destabilize the mRNA of TBC1D20, a critical autophagic modulator. This repression blocks autophagosome‒lysosome fusion, thereby compromising degradation efficiency [226]. Age-related cataracts (ARCs) are a degenerative eye disease. With age, the lens gradually becomes cloudy, resulting in decreased vision. Emerging evidence indicates that increasing or preserving autophagic flux diminishes lens epithelial apoptosis, thereby slowing both the initiation and progression of age-related cataracts. Noncoding RNAs (ncRNAs) serve as crucial modulators of autophagic processes in the pathogenesis of age-related cataracts. These RNA species employ various mechanisms to regulate target gene expression, primarily through microRNAs (mRNAs), long noncoding RNAs (lncRNAs) and circular RNAs (circRNAs) [227]. Experimental evidence revealed that the level of miR-23b-3p in lens epithelial cells increases; in this cell, miR-23b-3p inhibits SIRT1 and its downstream FOXO signaling pathway, eventually leading to a downward adjustment of the expression of ATG7 and BECN1 [228]. This cascade impairs autophagic elimination of abnormal proteins and compromised organelles, thereby accelerating cataract formation [228]. Let-7c-3p silences Atg3, an early-phase autophagy gene in lens epithelial cells, thereby suppressing oxidative-stress-driven autophagy and apoptosis [229]. Additionally, let-7c-5p modulates the ERCC6/VCP pathway, influencing the degradation phase of autophagy and participating in the pathogenesis of ARC [225]. These observations underpin a growing arsenal of drugs that tune autophagy for therapeutic gain. In aged-mouse models of cataract, low-dose metformin stimulates AMPK, boosts autophagic flow and slows lens opacity development [230]. Posterior capsular opacification (PCO), the commonest postsurgical sequela of cataract extraction, arises from proliferation, migration and epithelial-to-mesenchymal transition of residual lens epithelial cells within the capsule [231]. Effective prevention of PCO involves complete removal or inhibition of residual LECs during cataract extraction. Studies have shown that cyclosporin A, an immunosuppressant, can induce autophagic cell death in vitro, preventing PCO formation and highlighting the importance of autophagy in PCO prevention [232]. Similarly, when polyamidoamine (PAMAM) dendrimers are loaded into drug-eluting intraocular lenses, autophagy is enhanced, LC3-II levels increase, PCO is effectively inhibited after cataract surgery, and no damage to other ocular tissues is observed [233]. Epithelial-mesenchymal transition represents a fundamental process in posterior capsule opacification development after cataract surgery. Investigations demonstrate that surgical intervention significantly increases TGF-β2 concentrations in aqueous humor, serving as a primary inducer of EMT [234]. This cellular transition causes posterior capsule opacification, clinically termed fibrotic cataract [235]. Further research established that TGF-β2 stimulates LC3-I to LC3-II conversion and p62 degradation in lens epithelial cells, thereby promoting autophagic activity through enhanced autophagosome biogenesis and lysosomal fusion. Excessive autophagy therefore drives LECs into EMT and subsequently induces PCO [236]. Autophagy in retinal dystrophies Retinal dystrophies are genetically diverse disorders characterized by irreversible vision loss from photoreceptor failure and death [237]. Retinitis pigmentosa (RP), the most common subtype, begins with progressive rod loss that releases inflammatory mediators and ROS, disrupting photoreceptor homeostasis and ultimately endangering cone survival [238]. Retinal pigment degeneration is clinically manifested as retinal pigment deposition and irreversible vision loss, which is still a very challenging treatment problem. Genetic analysis shows that the disease can be caused by mutations in more than 80 different genes, many of which encode the regulatory elements in the visual signal transduction cascade [238, 239]. Previous studies have confirmed that physiological-level autophagy is crucial to the steady state of photoreceptors. Defective autophagy leads to intracellular accumulation of rhodopsin and transducin, precipitating photoreceptor demise (Fig. 5) [240, 241]. The rd10 murine strain, commonly employed as an experimental retinitis pigmentosa system, harbors a pathogenic variant in the photoreceptor-specific PDE6B β-subunit that severely compromises catalytic function. This molecular defect raises cGMP levels, initiating progressive rod photoreceptor demise followed by secondary cone degeneration [242]. Yang et al. showed that rd10 mice lacking the Sigma-1 receptor exhibit an early phase of retinal autophagy activation that protects rods. As rod death intensifies, massive intracellular Ca²⁺ accumulation switches the mode of death to apoptosis, while autophagic flux declines and cones succumb to necroptosis [243]. In multiple RP models, genes of the insulin/mTOR pathway are upregulated, indicating a state of nutrient stress; however, exogenous insulin lessens cell death, indicating that autophagy is recruited to counter the metabolic crisis [244]. Epigenetic studies further revealed that histone deacetylase inhibitors (e.g., TSA and tubastatin A) delay cone loss by enhancing autophagy and by reinforcing neuroprotective MAPK and PI3K/Akt signaling as well as the HIF-1 pathway [245, 246]. Autosomal dominant retinitis pigmentosa results from pathogenic variants in the rhodopsin gene, with the P23H substitution being most prevalent. This mutated visual pigment undergoes misfolding and aggregation within the endoplasmic reticulum, inducing significant endoplasmic reticulum stress [239, 247]​​. Although rapamycin administration consistently enhances scotopic electroretinogram responses, this protective effect occurs independently of autophagic protein elevation, instead involving unfolded protein response activation. Supporting this mechanism, Hsp70 overexpression similarly promotes photoreceptor viability in P23H-rhodopsin retinas [248]. It can therefore be inferred that in this model, the benefits of rapamycin may stem from its broad role in mitigating proteotoxic stress rather than from a specific enhancement of autophagy. However, excessive autophagy can, under certain conditions, exacerbate photoreceptor degeneration. Conversely, a measured restraint of autophagy may foster proteasomal activation, accelerate the disposal of misfolded rhodopsin and thereby promote photoreceptor survival. In a P23H-rhodopsin knock-in model, rod-specific deletion of Atg5 was found to improve both outer-segment ultrastructure and scotopic ERG responses [63], indicating that hyper-activated autophagy disrupts the fine balance between the autophagic and proteasomal pathways. In the rd10 murine model, increased cytosolic calcium levels before maximal rod degeneration subsequently activate calpain. Experimental evidence has demonstrated that pharmacological inhibition of calpain activity attenuates photoreceptor degeneration in both in vivo and in vitro models. These findings suggest that calpain-induced lysosomal membrane disruption contributes to lysosomal impairment and autophagic pathway suppression during retinal degeneration in retinitis pigmentosa [249]. This molecular mechanism aligns with the previously described pathway in glaucoma, where cathepsin leakage causes impaired autophagic flux and contributes to neurodegenerative processes in the optic nerve. Given that lysosomal injury represents a late-stage critical event, the most effective strategy should be to repair and protect lysosomes. For example, trehalose enhances lysosomal biogenesis and function by up-regulating the master regulator TFEB, and has demonstrated therapeutic efficacy across multiple models [249–251]. Autophagy is also indispensable for photopic vision. Across experimental systems where autophagic processes in rod photoreceptors are genetically modified, cone photoreceptor degeneration consistently emerges as a subsequent pathological outcome. Autophagy in retinal detachment Retinal detachment occurs when the neurosensory layer separates from the retinal pigment epithelium, depriving photoreceptors of metabolic support and inducing hypoxia and nutrient starvation that culminate in cell death [252]. Early after detachment, photoreceptors activate autophagy as a transient survival mechanism, yet prolonged detachment progressively impairs autophagic flux, accelerating their death (Fig. 5) [253]. Studies using the rodent retinal detachment model showed that the expression level of autophagy indicators (including Atg5, LC3-II and LC3-II/LC3-I ratios) increased and peaked on the third day after detachment. Within 24 h after induction detachment, the degree of apoptosis of the outer nuclear cell in the samples treated with autophagy inhibitor 3-MA was significantly higher than that of the untreated control group. In addition, rats transfected with siAtg5 showed reduced autophagy levels and increased Caspase-8 expression [253]. In order to clarify the exact upstream mechanism leading to this phenomenon, follow-up studies confirmed that retinal detachment can cause hypoxia and upregulate the expression of hypoxic-induced factors HIF-1α and HIF-2α. This oxygen deprivation increases BNIP3 expression via HIF-1α-dependent pathways, initiating mitochondrial autophagy in photoreceptors. Genetic suppression of HIF-1α diminishes LC3-II conversion and intensifies photoreceptor apoptosis [254]. Investigations using a murine retinal detachment system revealed that TNF-α levels significantly increased (4-fold increase) in photoreceptors within 24 h after induction but then progressively decreased. Parallel measurements revealed elevated LC3B-II/LC3B-I ratios and ATG5 expression during days 1–3 post-detachment, which normalized by day 7. Subsequent inhibition of TNF-α using infliximab delayed and moderated the peak expression of Atg5/LC3B, reduced the number of TUNEL-positive cells, and resulted in appropriately sustained autophagic activity until day 7 [255]. Nevertheless, separate investigations suggest light-induced retinal injury triggers autophagic processes that can paradoxically exacerbate damage, possibly through activation of non-apoptotic cell death mechanisms. Early studies have observed that in the process of loss of photosensitive cells, apoptosis-related genes and autophagy-related genes show co-expression. After peroxide stimulation treatment of 661 W cells, autophagy and apoptotic water increased on average, accompanied by an increase in cell mortality. Reverse transcription PCR results show that the transcription levels of LC3 and caspase-1 are upregulated, while the transcription levels of BIRC-4 are downregulated, thus confirming the synchronous upregulation of autophagy and apoptosis [256]. Inhibition of autophagy leads to reduced cell apoptosis and increased cell survival, which suggests that autophagy may trigger apoptosis in photoreceptors. Investigating the upstream regulatory factors of this process could offer novel therapeutic directions for treating photoreceptor pathologies [256]. Research by Kunchithapautham et al. revealed that extended rapamycin administration in a murine model of phototoxic retinal injury preserved rod viability, whereas cone photoreceptor functionality deteriorated [257]. Previous work has shown that suppressing endoplasmic reticulum stress-mediated autophagy provides protection against light-induced retinal damage [258]. These comprehensive research results show that under photostress conditions, autophagy regulation has completely different consequences in different photoreceptor subtypes. Additional investigations are needed to clarify the involvement of external and internal determinants in photoreceptor preservation. Autophagy in retinal ischemic injury Damage to eye perfusion will lead to ischemia and secondary tissue damage. When local circulation cannot meet metabolic needs, retinal ischemia occurs; this situation is usually caused by damage to local vascular function, not systemic hypoperfusion. This pathological state manifests in multiple ophthalmic disorders, such as anterior ischemic optic neuropathy, vascular occlusions, and microangiopathies, including diabetic retinopathy and glaucoma [259]. The inner nuclear layer neurons and retinal ganglion cells demonstrate particular susceptibility to ischemic injury [260]. Research reveals autophagy exhibits dynamic functionality in retinal ischemia, transitioning from cytoprotective mechanism to pro-death pathway under extreme metabolic stress (Fig. 6). Fig. 6Autophagy Mechanisms in Ocular Pathologies. This diagram depicts autophagy’s intricate contributions to ocular diseases such as retinal ischemic injury, lysosomal storage disorders, uveitis, corneal dystrophies, dry eye disease, and optic neuropathies. The analysis focuses on how impaired autophagic flux at key stages leads to compromised lysosomal activity, resulting in pathological outcomes including lipofuscin accumulation, protein aggregation, and inflammatory activation that collectively drive disease development. LHON, Leber’s Hereditary optic neuropathy; OPA1, optic atrophy 1; Atg, autophagy-related gene; NLRP3, NLR family pyrin domain containing 3; IL-1B, interleukin-1 beta; IL-18, interleukin-18; IL-17, interleukin-17; IL-1R, interleukin-1 receptor; ROS, reactive oxygen species; AMPK, AMP-activated protein kinase; DNA, deoxyribonucleic acid; RGCs, retinal ganglion cells; LMP, lysosomal membrane permeabilization; GM2, ganglioside GM2; MFN2, mitofusin 2; LAMP2, lysosomal associated membrane protein 2; TGFBI, transforming growth factor beta induced; CTSD, cathepsin D; FOXO, forkhead box O; DDIT4, DNA damage inducible transcript 4; ULK1, unc-51-like autophagy activating kinase 1 Autophagy Mechanisms in Ocular Pathologies. This diagram depicts autophagy’s intricate contributions to ocular diseases such as retinal ischemic injury, lysosomal storage disorders, uveitis, corneal dystrophies, dry eye disease, and optic neuropathies. The analysis focuses on how impaired autophagic flux at key stages leads to compromised lysosomal activity, resulting in pathological outcomes including lipofuscin accumulation, protein aggregation, and inflammatory activation that collectively drive disease development. LHON, Leber’s Hereditary optic neuropathy; OPA1, optic atrophy 1; Atg, autophagy-related gene; NLRP3, NLR family pyrin domain containing 3; IL-1B, interleukin-1 beta; IL-18, interleukin-18; IL-17, interleukin-17; IL-1R, interleukin-1 receptor; ROS, reactive oxygen species; AMPK, AMP-activated protein kinase; DNA, deoxyribonucleic acid; RGCs, retinal ganglion cells; LMP, lysosomal membrane permeabilization; GM2, ganglioside GM2; MFN2, mitofusin 2; LAMP2, lysosomal associated membrane protein 2; TGFBI, transforming growth factor beta induced; CTSD, cathepsin D; FOXO, forkhead box O; DDIT4, DNA damage inducible transcript 4; ULK1, unc-51-like autophagy activating kinase 1 In cerebral ischemia, autophagy exerts a lethal role; Koike et al. reported robust up-regulation of the pathway in a brain ischaemic model, and genetic deletion of the essential gene ATG7 protected neurons, providing strong evidence for a pro-death function [261]. Similarly, in an acute IOP-elevation model, increased autophagy markers were observed, and intravitreal injection of 3-MA, an autophagy inhibitor, protected retinal ganglion cells from hypoxia-induced death [262]. Conversely, autophagy activation with rapamycin enhanced apoptosis, confirming the pro-death effect in a reciprocal manner [263]. The underlying molecular mechanism involves intimate crosstalk between autophagy and necroptosis, another programmed necrotic pathway [264]. In a murine retinal ischemia system, programmed necrosis mediated cellular death within both the ganglion cell layer and the inner nuclear layer. Pharmacological suppression of programmed necrosis using Nec-1 not only diminished cellular death but also concurrently suppressed autophagic activity and promoted visual functional recovery [265], indicating that under ischemic conditions, autophagy activation may be part of the necroptotic program or that both pathways share common upstream signals. Although evidence is limited, autophagy may exert protective effects under specific conditions. In a rat model, ATG13 transcriptional activation occurred following ischemic preconditioning but remained unchanged during acute ischemic episodes, indicating that the induction of autophagy potentially represents a fundamental neuroprotective mechanism [266]. Notably, Calpain-mediated regulation of autophagy represents a potential molecular mechanism underlying the pathogenesis. Experimental models of ocular ischemia demonstrate that Calpain (a calcium-activated protease) is activated and cleaves Beclin-1, a key initiator of autophagy. This process can be inhibited by intravitreal injection of the NMDA receptor antagonist MK801 [184]. This cleavage is postulated to disrupt normal autophagic function, leading to dysfunctional autophagy and potentially promoting cell death. Beclin-1 is the core component of Class III PI3K complex, which is crucial for autophagosome biosynthesis. Calpain cleaves Beclin-1 near amino acid 149, generating an approximately 50 kDa N-terminal truncated fragment (ΔN-Beclin-1). This ΔN-Beclin-1 fragment exhibits enhanced pro-autophagic activity. It demonstrates a greater propensity to bind Atg14L, a positive regulator of autophagosome initiation, leading to the excessive and uncontrolled nucleation of autophagosomes [184, 267]​​. The absence of complete Beclin-1 destroys its binding to Vps34 and hinders the subsequent steps of autophagy maturation and lysosomal fusion, thus blocking the downstream autophagic flux​​ [184]. However, the current evidence mainly comes from rodent animal models, which fail to fully reproduce human retinal ischemic pathology, thus limiting their clinical transformation. Autophagy in lysosomal storage disorders Lysosomal storage disorders (LSDs) is a group of hereditary metabolic diseases caused by activity defects of specific lysosomal proteins or cofactors, resulting in the progressive deposition of lipids, glycoproteins or glycosaminoglycans. Ocular manifestations frequently involve retinal tissue, which presents with a characteristic macular appearance and neuronal degeneration due to intracellular sphingolipid accumulation [268]. Despite phenotypic diversity (e.g., neuronal ceroid lipofuscinosis, mucopolysaccharidosis, glycogen storage disease), LSDs share common features including impaired autophagy, ubiquitinated protein aggregation, mitochondrial dysfunction and inflammatory activation [269]. Although the underlying biochemical and genetic defects of most LSDs have been elucidated, the precise mechanisms leading to neuronal loss and visual impairment remain incompletely understood. Accumulating evidence indicates that autophagy pathway impairment is a shared hallmark of the majority of LSDs [270]. Consequently, disruption of autophagic flux is currently considered the key bridge linking lysosomal dysfunction to cellular pathology (Fig. 6). Owing to the accumulation of intraluminal debris and increased susceptibility to lysosomal membrane permeabilization (LMP), the terminal degradative step of autophagic flux is blocked. This prevents efficient fusion of autophagic vacuoles (e.g., autophagosomes) with dysfunctional lysosomes, resulting in the intracellular build-up of undegraded material [271, 272]. In addition, activation of the calpain–Beclin-1 axis, as mentioned above, contributes to the disruption of normal autophagy regulation [269]. In Gaucher and Tay‒Sachs disorders, accumulated glucosylceramide and GM2 ganglioside compromise lysosomal membrane integrity, triggering cathepsin B and L release. Such enzymatic activity facilitates NLRP3 inflammasome formation, resulting in the release of IL-1β and IL-18, exacerbating neural inflammation and compromising mitochondrial integrity, thereby establishing a self-sustaining pathological cycle [273]. Research employing animal models displaying diverse disease manifestations has clarified the mechanistic contributions of autophagy to retinal pathology. Investigations using Niemann-Pick type C1 (Npc1) deficient mice demonstrated compromised cone-mediated visual signaling through electroretinographic assessment. Morphological analysis showed sterol accumulation (Filipin-positive) beneath the ganglion cell layer (GCL) and lipofuscin-like inclusions in the RPE. Ultrastructurally, GCL cells exhibited vacuolar structures resembling multilamellar bodies or autophagosomes, along with strong LC3-II signal, clearly indicating disrupted autophagy [274]. In acid sphingomyelinase-deficient murine models of Niemann-Pick type A disease, electroretinographic recordings demonstrated diminished visual signal transmission during both scotopic and photopic testing. These functional alterations coincided with progressive structural alterations in retinal architecture and autofluorescent pigment deposition within retinal pigment epithelium. Immunoblot analysis revealed increased LC3-II conversion. Crucially, after lysosomal inhibitor treatment, the control cells showed further LC3-II accumulation (indicating that the autophagic flux function was normal). In contrast, the fibroblasts originating from the patient did not show such a reaction, which revealed the defects in the fusion of autophagosomes and lysosomes, and showed a typical autophagic flux blocking [154]. Pompe disease is caused by acidic alpha-glucosidase (GAA) deficiency, resulting in lysosomal glycogen accumulation [275]. The disease is characterized by significant LC3-II-labeled autophagy vesicle accumulation in muscle tissue due to damage to the autophagy process. Enzyme replacement therapy (ERT) has limited efficacy because the autophagy barrier hinders the delivery of enzymes [276]. On the contrary, inhibiting autophagy by genetic means (e.g., Atg7 knockout) can promote the clearance of glycogen [277, 278]. Neuronal ceroid lipofuscinoses constitute the primary etiology of pediatric neurodegenerative disorders, exhibiting diverse genetic origins and clinical manifestations [279]. The juvenile variant (Batten disease), the predominant subtype, stems from pathogenic CLN3 gene variants [280], while congenital forms arise from cathepsin D mutations (CTSD/CLN10) [281]. Notably elevated autophagic vesicle quantities appear in human subjects and murine models of these conditions [282]. CLN3 defects also provoke abnormal mitochondrial retention and suppressed mTOR signaling, indicating stimulated autophagosome biogenesis. The formation of autophagosomes increases but degradation is blocked, leading to neuronal death [282, 283]. In murine models of Sandhoff disease (GM1 gangliosidosis) with Hexb and β Gal deficiencies, although autophagic markers remained unexamined, ultrastructural analysis showed enlarged secondary lysosomes containing membranous lamellar inclusions alongside multiple vacuoles within ganglion cell layer neurons. These morphological characteristics strongly indicate autophagic impairment marked by defective clearance of autophagic cargo [284]. In summary, autophagy dysregulation constitutes a central pathogenic mechanism in LSDs: lysosomal defects impair autophagic flux, leading to toxic accumulation, energy crisis and cell death, thereby leading to a unifying pathogenesis for LSDs. Therapies should restore flux integrity (TFEB activation and lysosome stabilization) rather than globally modulating autophagy. Autophagy in uveitis Uveitis is an autoimmune-driven type of intraocular inflammation that is characterized by a paradoxical autophagy profile that varies with model type (Fig. 6). In a rat model of recurrent disease, increased autophagy within T cells correlated with increased retinal infiltration; rapamycin, an autophagy enhancer, exacerbated experimental autoimmune uveitis, suggesting a proinflammatory role that promotes T-cell activation, survival and migration, thereby amplifying the autoimmune assault [285]. Conversely, in a systemic/retinal inflammation model induced by intraperitoneal lipopolysaccharide (LPS) injection, rapamycin demonstrates a neuroprotective effect: it increases LC3-II levels and preserves visual function [286]. These divergent outcomes demonstrate that autophagic activity during inflammatory responses depends critically on both the specific inflammatory stimulus and the characteristics of participating immune cells. Investigations employing endotoxin-induced uveitis and experimental autoimmune uveitis models demonstrated that mice with tissue-specific Atg5 deletion developed exacerbated ocular inflammation. These animals exhibited 2.5-fold greater inflammatory cell recruitment into retinal vasculature and substantially increased circulating levels of inflammatory mediators including IL-6, IL-1β, and IL-10, while maintaining normal TLR4 expression [131]​​. This pattern shows that the enhancement of the inflammatory response is independent of the upregulation of the receptor. Parallel observations of other autophagy-related genes (Atg7, Atg16l1) confirm that the defects of autophagy treatment fundamentally exacerbate the inflammatory reaction, thus confirming that autophagy has a protective effect on the pathogenesis of uveitis. Analysis of inflammasome activity and proinflammatory cytokine relationships revealed that autophagy deficiency in macrophages (e.g., Atg5 deletion) leads to blocked autophagic flux, resulting in excessive inflammasome activation and caspase-dependent cleavage that promotes massive IL-1β secretion. High levels of IL-1β disrupt the blood‒retinal barrier and facilitate leukocyte infiltration and local inflammation, ultimately contributing to the development of uveitis [131]. MicroRNA analysis in rodents with experimental autoimmune uveitis revealed reduced miR-30b-5p expression. Lentiviral delivery to reinstate this microRNA downregulated the transcriptional activity of Atg5, Atg12, and Becn1; diminished autophagic vesicle quantities; elevated IL-10 production; reduced IL-17 secretion; restored the Th17/Treg equilibrium; and mitigated intraocular inflammation. These findings indicate that excessive autophagy can also exacerbate inflammation [287]. These insights substantially advance the mechanistic understanding of the pathogenesis of uveitis while revealing novel therapeutic strategies for immunologically specialized tissues. Potential interventions include blocking the IL-1β pathway using agents such as anakinra or implementing caspase suppression protocols. Autophagy in corneal dystrophies In the context of multiple corneal dystrophies, autophagy activation constitutes a well-documented cytoprotective mechanism (Fig. 6). Systemic lipopolysaccharides (LPS) can induce corneal damage, activate autophagy and promote nuclear transport of TFEB. TFEB is a key transcription factor that regulates lysosomal biological development and autophagic flux [288]. Granular corneal malnutrition type 2 (GCD2) is an autosomal dominant genetic disease characterized by extracellular deposition of transforming growth factor β-induced protein (TGFBIP). Corneal fibroblasts of GCD2 patients showed changes in autophagic flux, which was specifically reflected in the accumulation of p62 [288, 289]. Against this background, rapamycin and lithium can restore autophagic flux and enhance the degradation of mutant TGFBIp [290]. Thiel-Behnke corneal dystrophy (TBCD), resulting from TGFBI mutations, features aberrant protein deposition and subsequent corneal clouding. Familial analysis identified a heterozygous mutation (p.Gly623_Tyr626del) within the Fas1-4 domain, which disrupts the local hydrogen-bond network, converting stable β-sheet structures into flexible random coils and directly promoting protein aggregation. A mouse model generated via CRISPR/Cas9 technology to carry this identical mutation successfully recapitulated key human pathological features, confirming its pathogenicity [291]. Investigations into the mechanism of mutant TGFBI (mut-TGFBI) accumulation have shown that the protein colocalizes with LAMP2 but is not subject to altered ubiquitination, indicating that it is degraded via the autophagy–lysosome pathway rather than the ubiquitin–proteasome system. After impaired autophagosome‒lysosome fusion was ruled out, lysosomal dysfunction was investigated by LysoTracker staining, which revealed lysosomal alkalinization (loss of acidic environment), and proteomic analysis coupled with Western blotting confirmed the significant downregulation of the key hydrolase cathepsin D (CTSD) [291]. Concurrent alkalinization and CTSD deficiency act synergistically to compromise degradative capacity, causing a blockage in autophagic flux and subsequent accumulation of mut-TGFBI. In Schnyder corneal dystrophy, mutant UBIAD1, unable to undergo GGpp-mediated dissociation, is retained in the ER, evades autophagic degradation and drives cholesterol accumulation. Starvation or rapamycin-induced autophagy increases multilamellar bodies, whereas 3-MA-mediated inhibition reduces them, suggesting that autophagy here serves lipid sequestration rather than clearance [292]. Conversely, Fuchs endothelial corneal dystrophy exhibits ROS-driven, excessive mitophagy that degrades MFN2 and causes mitochondrial loss. Up-regulation of DRAM1/TMEM74 accompanies this overactive autophagy; lithium, via PI3K-dependent autophagy enhancement, improves endothelial cell survival, indicating that fine-tuning autophagy could represent a non-surgical therapeutic strategy for FECD [293]. Current animal models inadequately replicate human disease chronicity, whereas therapeutic modulation requires cell-type precision. Future work should focus on human tissue validation, targeted drug delivery, and the translation of mechanistic insights into personalized therapies to restore corneal homeostasis. Autophagy in dry eye disease Dry eye disease (DED) is a multifactorial surface disorder marked by inadequate tear volume or excessive evaporation; resultant tear-film instability and hyperosmolarity ignite inflammation that damages tissue and disrupts sensory function [294]. Under various stress conditions in dry eye disease, such as hyperosmolarity and desiccation, autophagy is typically activated as an adaptive response (Fig. 6). Enhancement of the autophagy/mitophagy pathway has been shown to alleviate inflammation and tissue damage [295, 296]. The ULK1 activator LYN-1604, which induces autophagy, improves tear secretion and reduces levels of inflammatory factors (TNF-α, MMP-3, MMP-9). Conversely, pharmacological inhibition of autophagy using 3-methyladenine (3-MA), a known phosphoinositide 3-kinase (PI3K) inhibitor [297, 298], aggravates the key pathological manifestations of dry eye disease. These conflicting effects confirm that autophagy, as an adaptive response, plays a protective role in DED. Specifically, the protective function of autophagy in dry eye disease is governed by a sophisticated regulatory network involving multiple critical signaling pathways. In addition to the aforementioned ULK1/mTOR initiation pathway, the DDIT4-ROS axis mediated oxidative stress pathways also play pivotal roles. Research by Wang et al. established that genetic suppression of DDIT4 via siRNA in corneal epithelial cultures diminished reactive oxygen species production during hyperosmotic conditions [299]. To evaluate the DDIT4-ROS pathway’s involvement in osmotic stress-induced autophagy, N-acetylcysteine (NAC) was administered. This antioxidant treatment normalized LC3-II and p62 concentrations in human corneal epithelial cells. Corresponding experiments in murine dry eye models showed topical NAC application preserved ocular surface integrity. These collective findings identify DDIT4 as an intrinsic modulator of redox-regulated autophagic activation [300]. In addition, representative studies on transcription regulation and homeostatic recovery pathways show that in the state of dry eye disease, the tear protein Lacritin can promote the retention and acetylation of transcription factor FOXO3 in the cytoplasm, thus correcting its abnormal nuclear localization. Then, this process activates autophagy and restores the homeostatic state of the corneal epithelium [301]. In parallel, investigations focusing on mitochondrial quality control pathways have revealed that melatonin and its drug delivery system significantly upregulate the expression of PINK1 and Parkin, while concurrently reducing p62 and TOM20 levels, leading to a marked enhancement of mitophagic flux. This mechanism alleviates ocular surface damage induced by hyperosmolar stress [302]. Research has demonstrated that the functional output of autophagy is highly context-dependent. For instance, in cellular models, neither TNF-α stimulation nor desiccation stress alone effectively induced autophagy. However, combined treatment with trehalose and TNF-α synergistically enhanced autophagic activity, as evidenced by an elevated LC3-II/I ratio, whereas co-administration of trehalose with desiccation stress failed to produce this effect. Correspondingly, in the desiccation-based model, trehalose significantly suppressed TNF-α-mediated inflammatory responses, yet only attenuated monocyte chemotactic protein-1 levels without markedly altering the concentrations of IL-6, IL-8, or MMP-9. These consistent findings from both in vitro studies and human trials indicate that the anti-inflammatory efficacy of autophagy is highly dependent on the specific type of stressor [303]. An additional study reported up-regulation of autophagy in corneal epithelial cells, evidenced by increased LC3B and the lysosomal marker LAMP1 [304]. Chloroquine-mediated inhibition of autophagy suppressed NF-κB activity without altering the autophagy level induced by dry-stress itself, implying the existence of an autophagy regulatory pathway independent of autophagosome–lysosome fusion. Collectively, these findings underscore the complexity of autophagy, whose effects appear to depend on the specific stress type and cellular context. Evidence from these investigations points to the regulatory roles of several key pathways in DED-related autophagy, which range from ULK1 and mTOR (rapamycin target) signaling to the DDIT4-ROS axis, FOXO3 acetylation, and PINK1/Parkin-driven mitophagy. This body of evidence robustly supports the targeting of autophagy pathways as an innovative therapeutic strategy for alleviating the pathological damage of dry eye disease. Autophagy in optic neuropathies Optic neuropathies typically present with acute or subacute symptom onset and cause focal rather than diffuse damage to the optic nerve or optic disc [305, 306]. Given that the neuroprotective role of autophagy is dependent on both cell type and stressor conditions, this section focuses on RGCs and their axons. It elucidates the fine regulatory mechanisms of the autophagic pathway under genetic and toxic pathological conditions, and how its dysregulation drives irreversible axonal degeneration and cell death. RGCs represent the principal cellular vulnerability in both dominant optic atrophy and Leber hereditary optic neuropathy. Although the underlying genetic alterations impair mitochondrial physiology throughout the organism, the pathological consequences nevertheless remain restricted to RGCs (Fig. 6) [307, 308]. Dominant optic atrophy (DOA) is chiefly caused by mutations in OPA1, whose protein product resides in the inner mitochondrial membrane and is essential for mitochondrial fusion and quality control [309]. Mutational spectra include splice variants, frameshifts, nonsense changes, deletions and duplications [310]. In heterozygous mice carrying OPA1 nonsense mutations, electron microscopy showed an increase in the autophagosome structures within RGCs [311]. Using the detection of LC3 spots overlapping with mitochondrial markers or lysosomal probes, it was confirmed that the mitochondrial clearance effect was significantly enhanced in fibroblasts from patients with autosomal dominant optic nerve atrophy with OPA1 defects [312, 313]. This process originates from AMPK signaling amplification triggered by OPA1 insufficiency, causing mitophagic vesicle aggregation in proximal axonal segments and consequent reduction of axonal mitochondrial content [314]. Leber’s hereditary optic neuropathy (LHON) results from pathogenic variants in mitochondrial DNA affecting respiratory chain complex I components. Frequently observed mutations involve ND1 (approximately 13%), ND4 (around 69%), and ND6 (nearly 14%) [315]. Cellular models expressing pathogenic ND4 or ND1 variants demonstrate impaired autophagic and mitochondrial clearance processes, leading to dysfunctional mitochondrial accumulation. Conversely, pharmacological induction of autophagy using rapamycin enhanced removal of compromised mitochondria, indicating therapeutic potential through enhanced autophagic activity [316, 317]. Similar to photoreceptors, RGCs are highly susceptible to mitochondrial dysfunction due to their long axonal structure and high energy demands [318]. In DOA and LHON, genetic mutations directly impair mitochondria, triggering an adaptive enhancement of PINK1/Parkin-mediated mitophagy [313]. This contrasts with the excessive autophagy-induced damage observed in photoreceptors during retinitis pigmentosa [63]. In the early stage of hereditary optic neuropathy, this enhanced mitochondrial autophagy mainly plays a compensatory protective role; and its final functional exhaustion may be closely related to the progression of the disease. In addition, ethambutol, a drug commonly used to treat tuberculosis, can induce optic neuropathy. Specifically, ethambutol disrupts the autophagy process by blocking autophagic flux and alkalizing lysosomes, eventually causing retinal ganglion cell apoptosis [319, 320]. Although poisoning and ischemia have fundamental differences in triggering factors and upstream signaling pathways, the drug-induced dysfunction of lysosomal degradation and the Calpain-mediated Beclin-1 cleavage that occurs in the later stage of retinal ischemic injury lead to the disorder of autophagosome maturation [184], ultimately resulting in the blockage of the autophagic flow and triggering similar neurodegenerative processes. Therapeutic strategies targeting autophagy in ophthalmic diseases In recent years, efforts to devise innovative therapeutics focused on autophagy modulation have intensified. In the following section, the latest research advances regarding the roles of various autophagy pathways in the treatment of ophthalmic diseases are summarized, the targets of different strategies (such as mTOR, TFEB, and noncoding RNAs) are analyzed, the advantages and limitations of various therapeutic approaches are compared, and their clinical translation potential along with the associated challenges are discussed. Regulation of the mTOR signaling pathway The mTOR signaling cascade functions as a master controller of metabolic processes. In terms of nutrient abundance, mTOR activation restrains autophagosome formation through phosphorylation-mediated suppression of ULK1/ATG13. Conversely, cellular stress inactivates mTORC1, permitting ULK1-initiated autophagy. In ophthalmic disorders, sustained mTOR signaling typically occurs, resulting in compromised cytoprotective autophagy and the aggregation of toxic proteins. Thus, targeting this pathway with inhibitors such as rapamycin to reactivate autophagy represents a promising therapeutic strategy. The first-generation drugs in this class are rapamycin and its analogs (rapalogs). Its specific mechanism of action involves the binding of FKBP12-rapamycin complex to the FRB domain of mTOR, thus inhibiting mTORC1 in a non-covalent way. Such drugs are characterized by high selectivity and relatively mild side effects; however, they can only partially inhibit mTORC1 (for example, its inhibitory effect on 4EBP1 is short-lived and ineffective), and usually cannot effectively inhibit mTORC2 [321]. Based on these findings, early drugs such as sirolimus and temsirolimus have been approved for immunosuppression and cancer treatment [322, 323]. Although blocking mTOR shows anti-cancer potential, rapalogs have failed to produce strong clinical efficacy; part of the reason is that inhibition of mTORC1 will eliminate the negative feedback effect on the PI3K-AKT pathway, thus amplifying the AKT intermediation. Lead the survival signal, which in turn offsets its anti-tumor effect [324]. FKBP12 is a peptidyl-prolyl isomerase whose catalytic activity is dispensable for rapamycin-induced growth suppression. The protein simply serves as an accessory factor in mTORC1 inhibition. Studies have shown that long-term administration of rapamycin (and by extension, rapalogs) leads to incomplete inhibition of mTORC1, such as failure to prevent the rephosphorylation of 4EBP1 [325–327]. To overcome these constraints, novel ATP-competitive mTOR inhibitors have been developed. These compounds directly target the catalytic site within the kinase region of mTOR, simultaneously blocking the activity of mTOR complex 1 and 2 [328]. Structural homology between the kinase domains of mTOR and PI3K enables certain agents to demonstrate dual targeting capability against both enzymatic pathways. By achieving more comprehensive and lasting inhibition of mTOR signal conduction, such inhibitors overcome the limitations of rapalogs and show excellent anti-tumor efficacy in preclinal models [328]. AZD8055 (AstraZeneca) exhibits strong mTOR suppression with 0.8 nM half-maximal inhibition, triggering autophagic processes and cellular demise. In the lung cancer xenotransplantation model, the drug can be effectively coordinated with the MEK inhibitor AZD6244, thus enhancing cytotoxicity and promoting tumor regression [329]. In addition, Palomid 529 [330], as a unique drug in pan-mTOR inhibitors, is currently being funded by the National Eye Institute of the United States to carry out clinical research on age-related macular degeneration. The test aims to use the anti-angiogenesis characteristics of mTOR inhibition to prevent neovascularization in the retina and subretinal cavity, Müller cell activation, and glial scar formation after retinal detachment [331]. In terms of cataract management, mTOR-targeting agents, including rapamycin, have demonstrated potential as pharmacological agents for slowing the progression of senility-e and diabetes-associated lens opacity. The influence of rapamycin on the viability of lens epithelial cells is conditional and cytoprotective, as determined by the cellular redox equilibrium. Research has indicated that under severe oxidative stress, autophagy inhibition may promote cell survival [332]. Hyperhomocysteinemia activates the mTOR pathway, suppresses autophagy, and reduces connexin expression. Rapamycin-mediated mTOR inhibition rescues these abnormalities, thereby slowing cataract development [333]. A study using a high glucose-induced rMC-1 cell model demonstrated that FTY720 significantly suppressed proinflammatory cytokine production by inhibiting mTOR/NF-κB signaling and modulating autophagy. These findings correlated with reduced rMC-1 cell injury, suggesting that FTY720 or related compounds targeting mTOR-associated pathways may serve as effective modulators for treating hyperglycemia-induced retinal damage [334]. The above research findings jointly establish the status of the mTOR pathway as the main inhibitor of autophagy, and reveal its important role in a variety of eye pathology processes. Although there are still many challenges in targeted delivery systems, personalized treatment plans and clinical transformation applications, rapamycin and its analogues have shown promising therapeutic potential by restoring the cell-protective autophagy process. The strategy based on the mTOR pathway will continue to promote the development of innovative and mechanism-oriented intervention methods for various ophthalmic diseases. Activation of transcription factor EB Transcription factor EB (TFEB) serves as the primary coordinator of the autophagic–lysosomal system. Unlike targets that affect only a single step of autophagy, TFEB transcriptionally activates the entire autophagy–lysosome gene network, thereby achieving synergistic enhancement of autophagic flux [335]. Ocular pathologies, including age-related macular degeneration, lens opacification, and glaucomatous neurodegeneration, are strongly correlated with TFEB impairment and aberrant protein deposition. Consequently, pharmacological or genetic potentiation of TFEB to reestablish cellular clearance capability has arisen as a strategically significant pathway-modifying therapeutic approach. During nutrient abundance, TFEB translocates to lysosomal surfaces where it undergoes mTORC1-mediated phosphorylation and cytoplasmic sequestration. Under cellular stress, TFEB dephosphorylation occurs, enabling nuclear migration and subsequent activation of downstream gene expression, thereby comprehensively augmenting degradation capacity [336]. Strategies targeting TFEB can be broadly divided into indirect and direct activation. Curcumin derivatives (e.g., C1) act as direct activators by directly interacting with TFEB, stabilizing its active conformation or promoting its nuclear translocation. These compounds can alleviate high-glucose-induced cataracts by clearing large amounts of oxidative stress products generated in the high-glucose environment, thereby reducing damage to lens epithelial cells [337]. Indirect activators are also widely used in clinical applications; for example, metformin can act as an AMPK activator to indirectly promote TFEB activation. In dry AMD, overactive AKT2 inhibits TFEB function; inhibiting AKT2 can restore TFEB activity via SIRT5, thereby rescuing lysosomal function and reducing the formation of pathogenic deposits (e.g., drusen) [338]. Another study revealed that the indirect activator trehalose can cause rapid and transient enlargement and permeabilization of the lysosomal membrane (LMP), inducing Ca²⁺ efflux and subsequent stimulation of the calcium-sensitive phosphatase PPP3CB/calcineurin. This enzyme subsequently mediates TFEB dephosphorylation and nuclear accumulation. By acting on a mechanism that does not depend on mTOR, the pathway promotes the removal of neurotoxic protein aggregates and shows neuroprotective effects in a variety of neurodegenerative disease models [339]. Regulation by non-coding RNAs Non-coding RNA species, particularly microRNAs, long non-coding RNAs and circular RNAs, function as essential gene expression regulators and play crucial roles in the fine-tuning of autophagic pathways. In ophthalmic diseases, the dysregulation of specific ncRNAs can profoundly disrupt the balance of autophagic flux by targeting core autophagy genes (e.g., Becn1, ATG5, mTOR), thereby contributing to disease pathogenesis. Consequently, restoring physiological autophagy levels using synthetically engineered ncRNA mimics or inhibitors (antagomirs) represents a highly promising therapeutic strategy characterized by enhanced specificity and reduced toxicity. This section will systematically explain the mechanism of the main ncRNA to regulate autophagy in eye lesions, and explore its clinical transformation prospects. Investigations reveal that oxidative stress substantially elevates miR-23b-3p expression in lens epithelial cells. Genetic suppression of this microRNA followed by hydrogen peroxide exposure led to substantially diminished apoptotic levels. Parallel analyses using immunostaining and immunoblotting techniques demonstrated elevated LC3-II and reduced p62 expression, confirming that miR-23b-3p inhibition augments cytoprotective autophagic activity in lens epithelium. Further work revealed that miR-23b-3p directly down-regulates SIRT1, thereby promoting apoptosis while inhibiting cytoprotective autophagy and accelerating cataract progression [228]. Long-chain non-coding RNA (lncRNAs) is a class of RNA transcripts that are more than 200 nucleotides long and do not encode proteins. They play a crucial regulatory role in a variety of biological processes such as cell differentiation, development and metabolic homeosis. In a relevant investigation, lens-like bodies (LBs) differentiated from human urine-derived induced pluripotent stem cells (UiPSCs) serve as a robust model for studying in vivo lens development, specifically the organelle degradation mediated by autophagy. Comparative transcriptome analysis shows that at multiple stages of development, LB shares a large number of differential expression genes with mouse lens [340]. Among these molecules, a long non-coding RNA associated with LC3B, designated ALB (Affecting LC3B), was identified as exhibiting high expression levels during lens fiber cell differentiation. Functional analyses established that ALB acts as a positive regulator of autophagic flux by promoting the lipidation of LC3B-I to its phosphatidylethanolamine-conjugated form, LC3B-II, thereby facilitating the clearance of organelles in the lens [340]. A number of ncRNAs have been shown to be involved in regulating the autophagy pathway in AMD. miR-34a, miR-184, and miR-146a indirectly influence autophagy and contribute to the pathogenesis of AMD by modulating microglial clearance via TREM2 regulation, RPE phagocytic function, and inflammatory responses, respectively, highlighting their potential as therapeutic targets [341, 342]. Currently, clinical trials of miRNA-based therapies are still in their nascent stages. For instance, Sirna-027 has been evaluated in clinical trials targeting VEGFR1 to validate the feasibility and safety of treating choroidal neovascularization secondary to AMD [343]. Next-generation agents, such as SYL1801 (administered as an eye drop formulation), are progressing through early-phase clinical trials, steadily advancing the clinical development of ocular RNA delivery [344]. Furthermore, the lncRNA NORAD plays a key role in maintaining the homeostasis of the RPE. Mechanistically, this involves its interaction with PGC-1α and SIRT1, leading to heightened acetylation of PGC-1α, elevated mitochondrial reactive oxygen species, and activation of the p53-p21 signaling axis, ultimately culminating in impaired mitophagy [345]. In DR, circMKLN1 is abnormally overexpressed in the serum of DR patients and in human retinal microvascular endothelial cells (hRMECs) treated with high glucose. circMKLN1 acts as a molecular sponge for miR-26a-5p, mediating hyperglycaemia/methylglyoxal-induced autophagy and influencing autophagic flux by regulating Rab11a expression. AAV2-mediated inhibition of circMKLN1 significantly improved capillary and vascular leakage in the acellular retinas of diabetic mice; this effect was reversed by rapamycin, confirming an autophagy-dependent mechanism [346]. These findings suggest that circMKLN1 may serve as a potential biomarker and therapeutic target for DR. Emerging evidence has highlighted the involvement of ncRNAs in the pathogenesis of glaucoma. In a study focusing on human Tenon’s capsule fibroblasts (HTFs), the long non-coding RNA NR_003923 was found to be upregulated in tissues from glaucoma patients. This lncRNA functions as a molecular sponge for miR-760 and miR-215-3p, thereby relieving their suppression of the downstream target IL22RA1. Consequently, it enhances TGF-β-induced proliferation, migration, fibrosis, and autophagy in HTFs. Knockdown of NR_003923 attenuates these pathological responses, suggesting its potential as a novel biomarker and therapeutic target for glaucoma [347]. Non-coding RNAs function as master regulators of autophagy, providing precise therapeutic targets for ocular disorders. Future work should focus on identifying new ncRNA biomarkers, developing an efficient ocular drug delivery system, and promoting candidate therapies from the preclinical stage to clinical trials, so as to establish ncRNA therapy as the next frontier in the field of ophthalmic precision medicine. Gene editing therapy Ophthalmic diseases, particularly those with a genetic component, have emerged as a frontline field for the clinical translation of gene editing because of their well-defined pathogenesis, the immune-privileged status of the eye, and the feasibility of local administration. The CRISPR platform has moved from a principle verification experiment to a human clinical trial, showing great application prospects in dealing with previously incurable blinding diseases. With ongoing technological iterations, gene therapy approaches have evolved from traditional CRISPR systems to more precise editing tools, outlining a clear developmental trajectory. The first-generation editing technology is the ​​CRISPR/Cas9 nuclease system​​. This system functions by creating DNA double-strand breaks (DSBs) to knock out pathogenic genes. For example, researchers used the CRISPR/Cas9 system mediated by the AAV vector to realize the genome editing of the VEGFA gene to target corneal neovascularization (CoNV). The intervention successfully prevented and treated suture-induced corneal neovascularization (CNV) in the mouse model, demonstrating its therapeutic potential [348]. CEP290 gene mutation is the cause of Leber congenital menigosis type 10 (LCA10). In the EDIT-101 clinical trial, the CRISPR/Cas9 system was delivered directly to the patient’s retina through subretinal injection to remove the pathogenic mutation in the intron. According to the test data, the vast majority of participants (about 79%) achieved measurable improvements in at least one of the four key visual and functional indicators. These included navigational ability, best-corrected visual acuity (BCVA), dark-adapted full-field stimulus testing (FST), and patient-reported quality of life. The promising safety profile of the intervention, which featured no serious adverse events, supports the preliminary effectiveness of the in vivo gene-editing strategy [349]. To mitigate the potential risks associated with double-strand breaks (DSBs), base editors have emerged. These tools enable direct and precise conversion of individual base pairs (e.g., A•T to G•C) without cleaving the DNA double strand, theoretically offering greater safety and efficiency. A pioneering study has achieved the first in vivo correction of a retinitis pigmentosa (RP)-causing SNV using an adenine base editor (ABE) delivered via adeno-associated virus (AAV), demonstrating high efficiency and a favorable safety profile [350]. The treatment successfully restored PDE6B protein expression and its correct localization to the outer segments of photoreceptor cells. H&E staining revealed that the treated area retained an outer nuclear layer 6–7 cells thick, and immunofluorescence showed the formation of structurally intact rod outer segments. By enabling precise in vivo gene editing, protein rescue, and structural preservation, this technology significantly ameliorates the disease phenotype in rd10 mice [350], presenting a promising and safer alternative for treating inherited retinal diseases caused by single-nucleotide mutations. Recently, investigators revealed HSV-1-erasing lentiviral particles (HELPs), a therapeutic platform that employs lentiviral vectors to deliver mRNA encoding CRISPR-Cas9 alongside gRNAs targeting vital HSV-1 genomic regions. It achieved, for the first time, efficient and safe eradication of the HSV-1 virus in vivo, not only by curing acute ocular infection but also by clearing the latent viral reservoir in the ​​trigeminal ganglion (TG)​​, thereby fundamentally preventing disease recurrence [351]. Its specific mechanism involves the delivery of Cas9 mRNA and gRNAs via HELPs. After infection, the translated Cas9 protein forms a complex with the gRNA, which specifically targets and cleaves the ​​UL8​​ and ​​UL29​​ genes within the HSV-1 genome, causing DSBs [351]. Due to the lack of efficient DNA repair mechanism, these fractures will lead to the comprehensive destruction and degradation of the viral genome, thus hindering the replication and assembly of the virus. This method confirms the feasibility of directly targeting the viral genome as a strategy for treating infectious diseases and provides an example for the development of CRISPR therapy for other viruses (e.g., HPV, HBV, and HIV) [352, 353]. Furthermore, ​​intrastromal injection is a well-established technique in ophthalmic clinical practice that facilitates clinical translation. Gene editing technology provides a revolutionary treatment prospect for ophthalmic diseases, especially hereditary blinding diseases with limited effectiveness of traditional therapy. Although challenges remain, including delivery efficiency and potential immune responses requiring further long-term assessment, with the continuous improvement of the delivery system and the continuous advancement of large-scale clinical trials, gene editing is expected to become an innovative treatment model in the field of ophthalmic precision medicine in the next decade. Artificial intelligence-assisted therapy Traditionally, the evaluation of autophagy activity at the tissue level has relied on invasive biopsies or in vitro cellular experiments. Artificial intelligence (AI), especially deep learning, is reshaping the diagnosis and treatment pattern of ophthalmic diseases by extracting complex features that cannot be detected by the human eye from massive medical images and clinical data. For instance, the DeepDT-Net network developed by Yoo et al., trained on color fundus images, can predict the outcomes of photodynamic therapy for central serous chorioretinopathy with an accuracy of 88%, providing an effective tool for retinal specialists to prognosticate treatment efficacy [354]. Similarly, in areas such as refractive and cataract surgery, AI is emerging as a crucial clinical decision-support tool by assisting in the screening of keratoconus and calculating intraocular lens (IOL) power [355]. Although artificial intelligence provides a new perspective for ophthalmic-assisted diagnosis and treatment, its application in autophagy-related eye diseases is still in the early stage. Currently, in the field of AMD, the DeepSeeNet framework developed by Peng et al. [356] has demonstrated the capability to identify reticular pseudodrusen and subretinal drusenoid deposits. Given that impaired autophagy flux in retinal pigment epithelial cells is a central event leading to lipofuscin accumulation and drusen formation, future efforts could leverage deep learning models like DeepSeeNet to quantify the volume and distribution of specific drusen subtypes and track subtle changes in RPE layer thickness. These image biomarkers derived from AI, as alternative markers of autophagy functional states, show broad application prospects in monitoring the progress of diseases. It is worth noting that although the research on the diagnosis and treatment of autophagy-related eye diseases of AI is still limited, its application in other medical fields has moved from the stage of concept verification to the stage of achieving substantial results. For example, Zhao et al. [357] developed the AIDPS model for prognostic prediction in prostate cancer by integrating autophagy-related genes from the HADb database and combining them with machine learning algorithms. Drawing on this concept, future efforts could integrate public ocular tissue transcriptome data with the HADb autophagy gene database, employing various machine learning algorithms to construct progression risk models for autophagy-related eye diseases. Furthermore, Beesabathuna et al. established a scalable image-based temporal phenotyping platform using AI technology, capable of accurately predicting biological changes following autophagy modulation through temporal morphological features [358]. Inspired by this, temporal deep learning models could be developed in ophthalmology to extract autophagy-related features from longitudinal imaging data, thereby predicting disease progression trajectories and treatment responses. In future, the convergence of AI and autophagy research is poised to advance to a deeper level. Integrating and analyzing patients’ ocular imaging and genomic data could enable AI to precisely identify specific patient subgroups whose disease progression is driven by impaired autophagy flux. For instance, before and after treatment with autophagy inducers such as rapamycin, AI could quantify imaging biomarkers like lens opacity and the efficiency of protein aggregate clearance as surrogate endpoints in novel drug trials. On the other hand, the AI-based platform can promote high-throughput screening of new compounds targeting autophagy pathways and predict their protective effect on retinal cells. This transformation research paradigm will strongly promote autophagy regulators from laboratory to clinical application, laying a solid foundation for precise ophthalmology based on autophagy regulation. Conclusion This review systematically outlines the dual, context-dependent roles of autophagy in ocular physiology and pathology. We emphasize that autophagy functions as an indispensable homeostatic regulator during the development, maintenance and degeneration of diverse ocular tissues, facilitating key processes ranging from organelle clearance in lens fiber cells and visual pigment recycling to retinal ganglion cell development and the modulation of ocular immune responses. Conversely, the dysregulation of autophagic activity, manifesting as either a blockade or hyperactivation of the flux, serves as a fundamental mechanism in the pathogenesis of a spectrum of significant blinding ophthalmic diseases, including age-related macular degeneration, glaucoma, diabetic retinopathy, and cataracts. Through comprehensive literature synthesis, we detail how autophagy dysfunction in different ocular compartments leads to pathological outcomes such as protein aggregation, mitochondrial impairment and inflammation. The discussion concludes by emphasizing a spectrum of therapeutic approaches aimed at the autophagy pathway, such as mTOR inhibitors, TFEB activators, non-coding RNA modulators, gene editing technologies, and AI-facilitated treatment strategies. This work provides a comprehensive synthesis of autophagy in ophthalmology, bridging fundamental molecular mechanisms with translational potential. It has considerable potential for developing interventions that enhance protective autophagy or suppress its cytotoxic forms. These span from established strategies, such as TFEB activation and noncoding RNA regulation to innovative approaches integrating gene editing and artificial intelligence-assisted diagnostics and therapeutics. The review’s strengths lie in its broad disease coverage and mechanistic insights from genetic models, though significant limitations remain. First, the reasoning and conclusions for some mechanisms rely heavily on animal models and immortalized cell lines, which often fail to fully recapitulate the chronicity, complexity, and cellular heterogeneity of human ocular diseases. Second, the fact that experimental outcomes are significantly influenced by factors such as the disease stage, cellular compartment (e.g., soma vs. axon), and metabolic milieu creates considerable complexity, posing challenges for analyzing the interrelationships between different autophagic pathways. Finally, a major limitation remains the absence of validated clinical assays for monitoring autophagic flux, impeding patient stratification and assessment of autophagy-targeted therapies in human trials. Future investigations should focus on advancing physiologically relevant human models, including patient-specific iPSC-derived retinal organoids and other ocular tissues, to strengthen the clinical applicability of preclinical findings. Furthermore, it is essential to refine dynamic monitoring techniques and apply multi-omics approaches to identify reliable biomarkers in accessible biofluids (e.g., aqueous humor or tears) for the precise assessment of autophagic flux in vivo. Regarding therapeutic translation, investigating the synergistic potential of combining autophagy modulators with existing standard-of-care treatments (e.g., anti-VEGF agents or neuroprotective drugs) could enhance efficacy and overcome resistance. The incorporation of artificial intelligence into clinical workflows holds promise for deciphering complex biomarker datasets related to autophagy and forecasting individual therapeutic outcomes, thereby facilitating enhanced diagnostic precision and personalized medicine. In summary, while technological progress has substantially advanced our understanding of ocular autophagy, research efforts must maintain a rapid pace of evolution. Addressing current challenges via profound, innovative, and clinically oriented investigations will be crucial to realizing the full potential of therapies targeting autophagy for ophthalmic conditions.