Cell Cleanup Processes at Nerve Connections and Other Areas: Links to Learning, Memory, and Brain Disorders

Autophagy, or self-eating, consists of several pathways that result in delivery of cellular constituents to the lysosome for degradation. Macroautophagy has been observed in all eukaryotic cells studied, including both single cell and multicellular organisms. It involves intracellular membrane extensions to form new organelles called autophagosomes that envelope intracellular cargoes. Autophagosome fusion with endocytic-lysosomal structures serve to deliver additional cargoes and degradative enzymes. Chaperone-mediated autophagy is observed in avian and mammalian cells, consisting of direct import of proteins with a KDEL-like motif into the lysosome in a process dependent upon HSPA8/Hsc70 chaperones and a specific isoform of the LAMP2 protein, LAMP2A. Microautophagy is a less understood process involving lysosomal membrane invaginations that sample nearby proteins. Unless otherwise stated, the term "autophagy" refers to macroautophagy in subsequent sections of this review.

It is important to note that the canonical pathway of mammalian autophagy induction was defined primary through studies of nutrient deprivation, in analogy to extensive work in yeast. Non-canonical pathways for initiating autophagosome formation represent additional emerging areas. For example, BECN1-independent autophagy/mitophagy was initially reported in the context of neuronal autophagy induced in response to mitochondrial injury [19]. In contrast to canonical extension of MAP1LC3-bearing membranes around the cargo from a single nucleation point [20], a live-imaging study in hepatocytes showed simultaneous recruitment of MAP1LC3 puncta to multiple spots on mitochondria damaged by photoirradiation, which subsequently fuse [21]. While the orientation of the recruited MAP1LC3 is unknown, possible mechanisms for direct recruitment of MAP1LC3 to the surface of damaged mitochondria include cardiolipin externalization for cargo recognition [22] and/or formation of phosphatidylethanolamine on the outer mitochondrial membrane [23]. The initiation of BECN1-independent autophagy remains poorly understood, but oxidative stress and mitogen-activated protein kinases have been implicated [19,24,25].

The brain is one of the most metabolically active tissues in the body, weighing in at only 2% of body mass, but consuming 20% of body energy [26]. In rodents, it has been estimated that the axonal compartment consumes 47% of brain energy, while the post-synaptic compartment consumes 34% [27]. In primates, the cortex is much thicker, reflecting increased complexities of signal processing, with functional MRI studies revealing that 75% of brain energy usage is post-synaptic [28]. Given the much greater efficiency of oxidative phosphorylation in extracting ATP from glucose, it is not surprising that neurons rely predominantly on mitochondrial respiration to sustain health and function [29,30]. As discussed in Part II, deficits in mitochondrial quality control have been linked to numerous neurological and neurodegenerative disorders.

Depending upon the nature of the stimulus, cargo or cell type, triggers for selective autophagy may recruit and activate the autophagy-initiating machinery through different pathways. Mitophagy induced by mitochondrial photoirradiation or complex I inhibition are mediated by mitogen-activated protein kinases, and do not require BECN1 [19,21]. Transmembrane mitophagy receptors such as BNIP3/NIX can directly recruit WIPI-ATG13 complexes, whereas FUNDC1 exclusively recruits ULK1 [45]. In depolarization-induced mitophagy, PRKN is not necessary for autophagy induction; rather BNIP3L/NIX inhibits MTOR signaling to activate canonical mitophagy [46]. PINK1-PRKN-dependent and a subset of PINK1-PRKN-independent mitophagy mechanisms are summarized in Table 1, and reviewed in depth elsewhere [17,47].

The entry and exit of proteins, nucleic acids, and organelles (such as synaptic vesicles and mitochondria) from the soma into the axon is regulated by molecular interactions in the axon initial segment [48]. Likewise, although the soma and dendritic compartments are often lumped together, specific mechanisms exist to promote or exclude cargos from entering the dendritic arbor. While most studies have concentrated on axonal trafficking, emerging knowledge suggests that distinct mechanisms govern dendritic transport [49].

A key question is whether autophagosomes in different neuronal compartments exhibit unique characteristics and regulatory mechanisms. Recent work indicates that during inhibitory avoidance learning, proteins involved in autophagy and lysosomal degradation are significant upregulated, primarily in neurons throughout the hippocampus and its subregions [50]. Notably, the autophagy and lysosomal proteins BECN1, MAP1LC3B and LAMP1 were elevated in both the soma and processes of neurons, whereas SQSTM1 increased only in the soma [50]. Additionally, SQSTM1 and LAMP1 showed higher levels in astrocytes and microglia [50]. The differential increase of these proteins in different cell types and compartments suggests a specific and compartmentalized regulation of autophagy-lysosomal activity in the brain during learning. Additional insights into the compartmentalization of autophagy within neurons have been provided by recent studies in primary neurons [51,52]. Autophagosomes originating in the axon quickly begin to mature, as indicated by the presence of the late endosome/lysosome marker LAMP1, and exhibit greater mobility compared to those derived from the somatodendritic compartment [52]. Interestingly, synaptic activity regulates the dynamics of autophagic organelles within neurons, particularly in dendrites as compared with axons [51]. Increased synaptic activity reduces the motility of autophagic organelles in dendrites and stimulates their maturation into degradative compartments, suggesting that these organelles are retained near active synaptic regions for localized degradation [51]. These findings suggest that the functional and structural compartmentalization observed in neurons may create molecularly distinct pools of autophagosomes, each tailored to meet specific functional demands. This compartmentalization contributes to the overall regulation of neuronal health and synaptic function.

Proper synaptic function demands active remodeling of the proteome in response to activity-based cues. Moreover, synaptic activity can create heavy demands on the local proteome, which can lead to damage. Thus, maintaining proteostasis at the synapse requires robust quality control pathways. Indeed, synaptic activity can stimulate the formation of autophagosomes in presynaptic terminals [57,66,72,73]. Moreover, reactive oxygen species (ROS)-induced damage to synaptic vesicle proteins is sufficient to rapidly stimulate autophagosome production in the presynaptic terminal [74]. These findings suggest a critical role for autophagy in regulating the synaptic proteome in response to activity or local damage.

But how is activity-based information decoded to generate an autophagosome? Local formation of autophagosomes in the distal axon involves core autophagy proteins conserved from yeast [54,75]. Interestingly, this process is regulated by proteins enriched in presynaptic terminals, including molecules involved in synaptic vesicle endocytosis [57,73], synaptic vesicle trafficking [76], and active zone proteins [77,78]. In Drosophila, Endophilin-A (EndoA; human homolog SH3GL2/endophilin A1), a BAR-domain protein required for synaptic vesicle endocytosis, may distinguish different signals that induce autophagy in presynaptic terminals, such as activity-induced calcium influx from metabolic stress [57] (Figure 2B). EndoA generates highly curved membranes that recruit the autophagy protein ATG3 to drive autophagosome formation [57]. This process is stimulated by phosphorylation of serine 75 in the amphipathic helix H1 of EndoA, which based on experiments with mammalian SH3GL2/endophilin A1 [79], is predicted to promote a shallower insertion of EndoA into membranes to generate curvature [57]. In fact, a non-phosphorylatable S75A mutant of EndoA blocks autophagosome formation induced by starvation [57]. A follow-up study identifies a residue (D265) in the disordered loop between the BAR and SH3 domains of EndoA that regulates EndoA flexibility and autophagosome formation in response to calcium influx, but not starvation; this mechanism is distinct from phosphorylation of helix H1 [73]. Thus, starvation and activity-induced calcium influx can promote autophagosome formation in presynaptic terminals via distinct mechanisms. Future studies will need to elucidate how different autophagy stimuli are coordinated with cargo selection.

The rate of autophagosome formation in the presynaptic terminal may be defined by the synaptic vesicle cycle in controlling the available pool of ATG9 (Figure 2B). ATG9 is a transmembrane lipid scramblase that drives autophagosome membrane expansion [80,81] (Figure 1B). It does so by flipping phospholipids, derived from a membrane source by ATG2 [82], between outer and inner leaflets of the phagophore membrane [80,81]. Indeed, depleting the pool of ATG9 in presynaptic terminals reduces autophagosome biogenesis in the axon [55,83,84]. ATG9-positive vesicles are distinct from synaptic vesicles [78,85,86], but they undergo cycles of endo/exocytosis that are coupled to synaptic activity and depend on the machinery that mediates synaptic vesicle cycling [e.g., the C. elegans DYN-1/dynamin 1/DNM1 homolog, UNC-26/synaptojanin 1/SYNJ1 homolog, and UNC-57/endophilin A/ SH3GL (1-3) homolog] [72]. Additionally, the C. elegans active zone protein CLA-1/clarinet may bridge exocytic active zones with endocytic periactive zones to regulate ATG9 sorting at synapses [78]. In this way, the neuron may coordinate autophagosome formation in presynaptic terminals with its activity state.

During development of the nervous system, neurons need to extend their axons to reach their correct targets to form functional synaptic connections [87]. Loss of autophagy genes can lead to defects in axon outgrowth and guidance, causing dysgenesis of interhemispheric tracts (bundles of axons that connect left and right hemispheres) in the murine brain [55,88–91]. In fact, a common pathology in congenital disorders of autophagy is a thinning of the corpus callosum [92]. Interestingly, BDNF (brain-derived neurotrophic factor), a growth factor in the central nervous system that supports neuronal development and survival, is an important regulator of neuronal autophagy [93–96]. BDNF can stimulate autophagy in axons [93,94]. Loss of autophagy in developing Drosophila photoreceptors leads to the formation of supernumerary synapses with aberrant neuronal partners [97]. In this model, autophagy may function to destabilize unfavored axonal filopodia (structures that survey synaptic partners) by degrading factors that promote assembly of the active zone (e.g., Liprin- and RhoGAP100F/Syd-1) [97]. Thus, autophagy functions early in synaptogenesis to restrict inappropriate synaptic connections. In another neuronal subtype in the fly visual system, the dorsal cluster neurons, autophagy may need to be suppressed during later stages of synaptogenesis to prevent aberrant degradation of active zone proteins and loss of mature synapses [98]. In short, the roles for autophagy in the axon are dynamic and depend on neuronal subtype and context, and involve the degradation of different cargoes. Further evidence of neuron subtype-specific roles for autophagy has also been revealed in C. elegans deficient for atg9 [55]. This study showed that autophagy is required for early stages in axonal outgrowth of nociceptive sensory PVD neurons and for the organization of presynaptic terminals in AIY interneurons. In total, compelling emerging evidence indicates key roles for autophagy in establishing proper connectivity of the nervous system.

Knockout of key autophagy genes can also lead to alterations in synaptic function. Indeed, deletion of Atg5 in excitatory neurons in the cortex and hippocampus increases presynaptic neurotransmission [99]. What is the mechanism? Several groups find that neurons deficient for key autophagy genes (e.g., Atg5 or Atg7) exhibit an accumulation of tubular endoplasmic reticulum (ER), but only in axons and not in dendrites [99,100]. This accumulation of ER can lead to elevated calcium release from the ER via ryanodine receptors, triggering secretion of synaptic vesicles [99]. Thus, a key function for axonal autophagy is to maintain axonal ER and synaptic calcium homeostasis to prevent aberrant neurotransmission (Figure 2B).

Autophagy may also constrain presynaptic neurotransmission by regulating the synaptic vesicle pool. Pharmacological stimulation of autophagy can dampen the synaptic vesicle pool size [101,102]. Conversely, loss of autophagy elicits an increase in synaptic vesicle release probability [102] and evoked neurotransmitter release [101].

Are synaptic vesicles or their components substrates for autophagy? Synaptic vesicle proteins are present in a subset of autophagosomes as assessed using immunofluorescence [102]. Another study finds that deletion of BSN/Bassoon, a presynaptic scaffolding protein of the active zone that also functions as a negative regulator of autophagy by sequestering ATG5 [77], increases the colocalization of synaptic vesicle proteins with autophagosomes and decreases the synaptic vesicle pool size [103] (Figure 2B). Bassoon negatively regulates PRKN-dependent ubiquitination of synaptic vesicle proteins, thereby controlling their rate of clearance by autophagy [103]. Insights gleaned from proteomic analysis of autophagic organelles isolated from mouse brain have identified presynaptic content that includes synaptic vesicle proteins, proteins involved in synaptic vesicle trafficking, and components of the active zone [104,105]. Interestingly, synaptic cargos are enriched in adult and aged mice [104]. Autophagy can also selectively degrade individual synaptic vesicle proteins that are focally damaged [74], rather than degrading the entire synaptic vesicle. Failure to remove damaged synaptic vesicle proteins by autophagy dampened excitatory postsynaptic current amplitudes [74], suggesting important roles for autophagy in protecting the integrity of the synaptic vesicle proteome for proper neurotransmission. However, other reports in autophagy-deficient neurons do not find alterations in synaptic vesicle number [100] or bulk levels of synaptic vesicle proteins [99,106]. Future studies are needed to elucidate how different autophagy stimuli are coordinated with cargo selection.

An intriguing possibility is whether autophagosomes might be secreted from presynaptic terminals. Proteins that affect synaptic development versus activity-induced synaptic remodeling of boutons at the neuromuscular junction were identified using a comparative RNAi screen in Drosophila [107]. Knockdown of core autophagy proteins involved in autophagosome formation (e.g., Atg1 or Atg8) disrupted both synapse formation and activity-induced synaptic remodeling [107]. Interestingly, proteins that selectively disrupted synapse development were highly enriched in the lysosomal pathway and included lysosomal degradative enzymes and lysosomal membrane proteins [107]. By contrast, proteins that selectively disrupted activity-induced synaptic remodeling were enriched for modulators of autophagy activation rather than degradation [107]. In fact, neuronal activity stimulated the production of autophagosomes in synaptic boutons but decreased their colocalization with lysosomes [107], suggesting a detour from the canonical fate of degradation. Proteins that disrupted activity-induced synaptic remodeling also included Snap29 [107], a SNARE implicated in secretory autophagy [108,109]. Presynaptic knockdown of the machinery involved in secretory autophagy (e.g., Sec22 and Snap29) impaired activity-based synaptic remodeling but not synaptic development [107]. Inhibiting autophagosome secretion prevented activity-induced increases in bouton size and postsynaptic AMPA receptor GluR (human homolog GRIA) levels, altering synaptic plasticity [107]. These data suggest that autophagosomes may have distinct fates, degradative versus secretory, and synaptic activity may drive autophagosomes toward the secretory route. This outcome, however, may be dependent on the type and duration of the activity stimulus as more sustained activity seems to drive degradative functions, as discussed above.

What might be the functional advantage of secreting autophagosomes? Secretory autophagosomes may provide membrane to enable synapse enlargement in response to activity. It has been proposed that cellular stress stimulates secretory autophagy to release factors that promote maturation of BDNF [110]. Secretory autophagy may also provide an alternative mechanism to dispose of cellular waste, particularly in the context of lysosomal dysfunction [111–113]. It will be interesting to further resolve the function of secretory autophagosomes at synapses moving forward.

The regulation of synaptic growth and plasticity is critical for the proper development of neural circuits that govern behavior and their ability to adapt to experiences and environmental changes in processes related to learning and memory. Postnatal synaptic development is a dynamic process that involves both the formation and the elimination or pruning of synapses [114,115]. These dynamic changes are crucial for the selection and maturation of synapses and neural circuits. While relatively little is known about the underlying mechanisms by which neuronal autophagy influences synaptic structure, earlier studies in Drosophila demonstrate that a developmental loss of autophagy reduces the size of the neuromuscular junction. Conversely, enhancing neuronal autophagy by overexpression of Atg1, a homolog of ULK1, results in increased numbers of synaptic sites [116]. However, in the mammalian brain, autophagy is inversely correlated with synaptic numbers. One study suggests that autophagy is essential for developmental pruning of dendritic spines, a process that appears to be impaired in individuals with autism spectrum disorder (ASD) [11]. The authors observed an elevated density of dendritic spines, increased levels of DLG4/PSD95 (discs large MAGUK scaffold protein 4), and reduced developmental pruning of these spines in layer V pyramidal neurons within the temporal lobe of postmortem ASD brains. Interestingly, dendritic spine pruning deficits in ASD correlates with hyperactivated MTOR and impaired autophagy, with a statistically significant increase in phosphorylated MTOR in the brains of ASD patients compared to age-matched controls [11]. Consistently, ASD patients showed significantly lower levels of MAP1LC3-II, an autophagosome marker, throughout childhood and adolescence [11]. Given that MTOR activation suppresses autophagy, this heightened MTOR activity suggests a reduction in autophagy levels [117]. Additionally, transgenic mice with hyperactivated MTOR displayed ASD-like behaviors and impaired spine pruning in cortical projection neurons, which were corrected by administering the MTOR inhibitor rapamycin [11]. Collectively, these studies emphasize a critical role for autophagy in synaptic pruning and suggest that disruptions in autophagy may be linked to ASD phenotypes.

The mitophagy protein PRKN, which is associated with autosomal recessive PD, is also involved in regulating post-synaptic pruning [12] (Figure 2C). Knockdown of PRKN resulting in a proliferation of glutamatergic synapses and increased vulnerability to excitotoxic injuries, indicating that one physiologic role for PRKN involves downregulating excitatory synapses. Interestingly, PRKN interacts with the post-synaptic density and regulates the degradation of glutamate receptor subunits (Figure 2C) [118,119]. It also promotes degradation of SYT11 (synaptotagmin 11), a putative PD risk factor [120]. SYT11 not only regulates the presynaptic vesicle recycling pathway, but also regulates anterograde and retrograde trafficking of distinct vesicles and their distributions throughout the soma, axon and dendrites of neurons. PRKN interacts with SH3GL/endophilin-A in a phosphorylation-sensitive manner to ubiquitinate synaptic protein complexes [121]. However, the relative roles of proteasomal versus autophagic degradation in relation to synaptic regulation downstream of PRKN would need to be experimentally determined.

Synaptic plasticity is defined as the activity-dependent strengthening or weakening of synaptic transmission over time, primarily through modifications to postsynaptic receptors on neurons. These include glutamate receptors that bind N-methyl-D-aspartate (NMDA, composed of GRIN/GluN subunits), alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA, composed of GRIA/GluA subunits), and kainate (composed of GRIK/GluR subunits) (Figure 2C). The ability of autophagy to degrade postsynaptic scaffolding proteins provides another potential mechanism by which it can regulate synapse structure and influence the efficacy of synaptic transmission (Figure 2C). During synaptogenesis, when presynaptic terminals engage with the postsynaptic cell, the postsynaptic cell upregulates local protein synthesis, accompanied by clustering and anchoring of neurotransmitter receptors and scaffolding molecules at the postsynaptic membranes [122–124]. Following patterned neuronal activity, postsynaptic proteins are either incorporated into or removed from the synapse, leading to enduring changes in synaptic strength [123,124].

Previous research involving C. elegans show that autophagy also facilitates the degradation of inhibitory receptors from postsynaptic and non-synaptic regions. The neurotransmitter gamma-aminobutyric acid (GABA) mediates inhibitory neurotransmission through ligation of fast-acting inotropic GABAA receptors and G-protein coupled GABAB receptors. In muscle cells, the clustering of postsynaptic GABAA receptors relies on presynaptic input [125]. GABAA receptors on postsynaptic spines that fail to receive input from GABAergic terminals are internalized for autophagic degradation [126]. Notably, this autophagic degradation is selective to GABAA receptors, with acetylcholine receptors in the same cells being spared [126]. These data indicate that transmembrane postsynaptic receptors can be selectively degraded via autophagy in an endocytosis-dependent manner.

In addition to involvement in developmental synaptogenesis, autophagy pathways are triggered by synaptic activity to selectively degrade postsynaptic proteins, thereby modulating the long-term synaptic remodeling necessary for learning and memory. Either high potassium stimulation or chemically induced long-term depression (LTD) briefly increases the number of autophagosomes and the levels of autophagy-related proteins in dendritic shafts and spines of rat hippocampal neurons [127]. Furthermore, the levels of the AMPA receptor subunit GRIA1/GluR1 decreased following chemical LTD, with this degradation being partially reduced by autophagy inhibitors [127]. These findings indicate that postsynaptic autophagy, regulated by neuronal activity, is at least partially responsible for the receptor degradation required for maintenance of LTD. Later research provides further evidence that autophagy modulates synaptic biology by directly degrading synaptic proteins [95]. In the absence of autophagy in cortical pyramidal neurons, levels of three prominent scaffold proteins that are essential for dendritic spine remodeling are elevated: DLG4/PSD95, PICK1, and SHANK3 [95]. By isolating and purifying autophagosomes from the mouse brain, this study showed that synaptic proteins are present inside autophagosomes and serve as direct substrates of autophagy [95].

Synapses serve as crucial communication sites between neurons and require significant energy to support essential molecular and cellular processes [128]. Mitochondria are vital organelles at these sites, providing the energy to establish synaptic circuitry and promote synaptic plasticity [128,129]. Mitophagy is a form of selective autophagy involved in the removal of damaged mitochondria [47,130]. PINK1 is a serine/threonine kinase that has been extensively explored for its roles in regulating mitochondrial respiration, fission-fusion, mitophagy, and transport [47,130–132]. Interestingly, significant cognitive impairments are frequently observed among human PINK1 mutation carriers [133–136], and PINK1 plays an important role in promoting dendritic branching, synaptic spine density in vitro and in vivo, and spine maturation [13,137].

When mitochondria are depolarized, PINK1 accumulates on the outer mitochondrial membrane, and upon autophosphorylation, recruits the E3 ubiquitin ligase PRKN [47,130]. PRKN ubiquitinates outer mitochondrial membrane proteins, tagging the mitochondria for degradation [47,130]. Several earlier studies [138,139] suggest that deficient mitophagy may contribute to synaptic dysfunction. However, more recent studies show that mice lacking PINK1 or PRKN do not spontaneously develop any overt phenotype, and the basal levels of mitophagy in the brain remain unchanged despite the absence of these proteins [140]. In contrast, PINK1 and PRKN are essential for hypoxia-induced mitophagy in larval wing disc tissues [141], indicating tissue/cell type specificity. Primary neurons exhibit diminished PRKN-mediated mitophagy responses compared to proliferative tumor cells [142,143]. Moreover, both in vitro and in vivo studies reveal that neurons engage in PINK1-PRKN-independent mitophagy (e.g., the cardiolipin and MARCHF5 pathways) [22,39,140,144] (Table 1). A developmental absence of the autophagy protein ATG5 causes only a transient delay in the acquisition of normal hippocampal spine densities [145]. In contrast, pink1 knockout elicits persisting spine deficits in both embryonic and adult mouse neurons at 6 months of age [13], implicating additional mechanisms. There are also mitophagy-independent pathways to remove mitochondria from axons [146]. Given these observations, and incomplete knowledge concerning the spectrum of mitophagy pathways engaged in primary neurons, the roles of specific mitophagy pathways in regulating post-synaptic health and function remain unclear.

What other mechanisms might be engaged by PINK1 and/or PRKN in support of synaptic health and function? PINK1 and PRKN work together not only in mitophagy, but also in regulating mitochondrial biogenesis [147–149]. Both proteins regulate mitochondrial dynamics, albeit through different mechanisms [150–152]. PRKN localizes to the postsynaptic density, where it mediates degradation of receptors [118,119] (Figure 2C). It also regulates vesicular transport or recycling by suppressing SYT11 expression [120]. PINK1 regulates post-synaptic mitochondrial calcium fluxes [153] and participates in a cytosolic pathway involving NSFL1C/p47 to promote dendritogenesis [137] and the maturation of dendritic mushroom spines [13].

Mitochondria undergo dynamic morphological changes and actively traffic within neurons. Changes in mitochondrial morphology and distribution regulate energy supply, calcium homeostasis, and other fundamental aspects of neuron structure and physiology [129]. Dendritic mitochondrial dynamics contribute to the induction of GRIN/NMDA receptor-dependent long-term potentiation (LTP) [154], the maintenance of long-term plasticity [155], and the facilitation of plasticity-induced local protein translation [156]. In particular, a burst of mitochondrial fission is required for LTP [154]. Interestingly, PINK1 can phosphorylate DNM1L/Drp1 at S616 to promote mitochondrial fission [157]. While it may be tempting to conclude that PINK1 supports synaptic development and plasticity by promoting mitochondrial fission (Figure 2C) [150,158,159], in mammalian neurons, PINK1 also acts to promote mitochondrial elongation [151]. Moreover, recruitment of endogenous PRKA/PKA to mitochondria rescues multiple PINK1-deficient phenotypes in primary neurons through phosphorylation of DNM1L/Drp1 at a site that suppresses fission (S656-rat; S637-human) [160]. Fission related to mitophagy or cell death is regulated differently from fission related to mitochondrial proliferation [161]. The regulation of dynamic mitochondrial fusion-fission cycles in the setting of synaptic activity, and the potential role of PINK1, represent areas for future study.

In axons, PINK1 phosphorylates Miro1, an outer mitochondrial membrane-associated Rho GTPase involved in mitochondrial transport. Depending upon the site of PINK1-linked phosphorylation, RHOT1/MIRO1 is either stabilized or undergoes PRKN-dependent proteasomal degradation. Stabilization is mediated by PINK1-linked phosphorylation at threonine 298/299, while degradation is mediated by phosphorylation of serine 156 [162,163]. Given the essential role of mitochondria in providing energy and buffering calcium at excitatory synapses [164,165], PINK1 may stabilize mitochondria in specific areas with high energy requirements to support synaptic development, while also acting to promote mitochondrial transport into dendrites [132,166].

Another mechanism by which PINK1 regulates synaptic mitochondria is through its ability to regulate mitochondrial calcium handling. Synaptic activity causes massive cytosolic calcium fluxes, buffered by rapid mitochondrial calcium uptake through the mitochondrial calcium uniporter, followed by a gradual release back to the cytosol. The proper release of mitochondrial calcium is essential for some forms of synaptic potentiation [167], and deficient release has been implicated in mitochondrial calcium overload [2,168]. An early study showed that PINK1 loss of function impairs the release of mitochondrial calcium to the cytosol [169]. Subsequently, a novel phosphorylation site of the mitochondrial calcium antiporter SLC8B1/NCLX was discovered, which is regulated by PINK1 via PRKA/PKA [153]. An SLC8B1^S258D phosphomimic completely rescues defective mitochondrial calcium release in dopaminergic neurons from pink1 knockout mice and protected them from dopamine-induced cell death, indicating the importance of proper mitochondrial calcium handling in neurons [153].

Mitochondrial calcium overload triggers the selective loss of post-synaptic mitochondria via mitophagy during excitatory stress elicited by mutations in LRRK2 (leucine rich repeat kinase 2) [170,171]. Enhancing mitochondrial calcium egress by mimicking the PINK1-regulated SLC8B1/NCLX phosphorylation at serine 258 confers protection against postsynaptic degeneration in this model. While the exact role of calcium-regulated mitophagy and altered mitochondrial dynamics remain to be fully elucidated in relation to excitatory neurotransmission [2], it is clear that enhancing PINK1 expression improves structural and functional parameters implicated in learning and memory [137,172–174].

PINK1 primarily localizes with mitochondria, but is also released into the cytosol following processing by mitochondrial peptidases for signaling or clearance [151,175,176]. Recently, cytosolic PINK1 was found to mediate mitophagy-independent signaling pathways important for maintaining the dendritic arbor [166,177]. The loss of endogenous PINK1 leads to dendritic simplification and synaptic deficits in cortical, hippocampal and midbrain neurons [13,150,166]. Conversely, upregulation of PINK1 promotes dendritic morphogenesis and synaptic maturation [137,174]. To determine which subcellular pools of PINK1 are responsible for promoting neurite extension, cells were transfected with two distinct PINK1 constructs: one that targeted PINK1 to the outer mitochondrial membrane (OMM-PINK1) and another that restricted PINK1 to the cytosol (ΔN111-PINK1) [166]. Both constructs prevented cell death due to the loss of endogenous PINK1. However, transfection with the cytosolic form (ΔN111-PINK1), but not the mitochondrial form (OMM-PINK1), significantly promoted dendritic growth in primary neurons [166].

How does cytosolic PINK1 regulate dendritic remodeling? Recent studies indicate that PINK1 interacts with valosin-containing protein (VCP) [137,178]. In the cytosol, this interaction scaffolds the activation of PRKA/PKA to phosphorylate NSFL1C/p47, a cofactor of valosin-containing protein that regulates membrane dynamics, at serine 176 [137]. A NSFL1C/p47^S176D phosphomimic effectively restores normal dendritic complexity and spine densities in neurons lacking PINK1 expression (Figure 2C) [13,137]. Interestingly, the yeast homolog of NSFL1C/p47, Shp1, is essential for autophagosome biogenesis, and co-expression of NSFL1C/p47 can ameliorate the autophagic blockage caused by VCP mutations [179,180]. The potential role of PINK1-regulated NSFL1C/p47 phosphorylation on autophagy, and the relative importance of autophagic versus non-autophagic mechanisms in mediating the effects of PINK1 on spine structure and function remain to be defined through future studies.

As discussed above, autophagosomes are actively assembled at synapses and along neuronal processes in hippocampal neurons [181–183]. Further, mitophagy-linked proteins such as PINK1 and PRKN regulate several aspects of post-synaptic structure and function. The well-documented age-related decline in hippocampal-dependent memory is correlated with a reduction of autophagy proteins in the brain, wherein ATG5, ATG7, and BECN1 are downregulated [184–186]. These observations implicate a role for autophagy in learning and memory. What then is the impact of autophagy modulation on specific learning behaviors?

The involvement of autophagy in memory formation, retrieval, and learning has been extensively documented in several recent studies [187–189]. Pharmacological inhibition of autophagy by infusing spautin-1 in the hippocampus before water maze training impairs long-term memory retention in mice. Conversely, activation of autophagy with the Tat-Beclin1 activator peptide enhanced long-term memory [189]. Subsequently, inhibiting autophagy by viral-mediated knockdown of autophagy proteins RB1CC1/FIP200, BECN1, or ATG12, or hippocampal administration of spautin-1, impaired novel object recognition and contextual fear conditioning. In contrast, administration of the autophagy activator Tat-Beclin1 enhanced retention of both types of memories [188]. Intriguingly, autophagy activity declines in the hippocampus with aging. Fully restoring autophagy levels is both necessary and sufficient to reverse associative and spatial memory deficits in aged mice [188].

The role of autophagy in brain function has also been investigated through cell-type-specific knockouts of autophagy genes. Mice with conditional knockout of Atg7 in the two main types of spiny projection neurons in the striatum illustrate a role for autophagy in uniquely regulating neuronal activity in different cell types involved in action selection and reinforcement learning [190]. Autophagy is essential for maintaining normal dendritic structure and synaptic input in spiny projection neurons of the direct pathway within the basal ganglia [190]. Interestingly, the loss of ATG7 in spiny projection neurons of the indirect pathway did not alter dendritic complexity, spine density, or excitatory inputs, but caused intrinsic hyperexcitability due to reduced Kir2 function [190]. Despite the differing cellular effects observed in neurons of the direct and indirect pathways, the loss of ATG7 in either neuron type leads to deficits in striatal-based behaviors [190]. This suggests that autophagy is crucial for both types of neurons in maintaining normal brain function and behavior, albeit through distinct mechanisms.

While autophagy clearly plays a role in memory formation, emerging evidence suggests that autophagy is not essential or beneficial for all aspects of learning and memory. For example, autophagic vesicles form locally in dendrites following stimuli that produce long-term depression (LTD) [14], a major form of long-lasting synaptic plasticity that supports cognitive function. These vesicles degrade GRIA/AMPAR subunits and other critical postsynaptic proteins. Unexpectedly, mice with conditional autophagy deficiency in excitatory neurons exhibited enhanced cognitive flexibility, outperforming controls in related behavioral tasks [14]. Similarly, while reduced autophagy impairs certain cognitive functions, preserved locomotion indicates that not all neural circuits are equally affected [191]. It is possible that too little or too much autophagy at the synapse, or the nature of the cargo being degraded, affects the outcome of autophagy modulation, just as the impact of autophagy on neuronal cell death or survival is context dependent [192]. The relationship of synaptic autophagy to synaptic plasticity, and the mechanisms by which autophagy is regulated in the context of learning and memory remain incompletely understood.

ATG9 is an essential protein involved in autophagy, the only known mammalian ATG protein to have a transmembrane domain. ATG9 participates in the formation and transport of vesicles that are crucial for the assembly of the autophagy machinery. These vesicles serve as platforms for autophagosome formation, providing the necessary lipids and proteins for the expansion of the phagophore, the precursor to the autophagosome [193]. In mammals two homologs of ATG9 are expressed: ATG9A and ATG9B. Atg9A-deficient mice die during embryonic development [194] or within one day of birth with a similar phenotype as Atg5-, Atg7-, and Atg16L1-deficient mice [195]. Conditional knockout (CKO) specifically targeting the central nervous system (CNS), causes axon-specific lesions like those observed in atg7-CKO mice [91]. Additionally, atg9a-CKO mice showed growth retardation, indicating the critical role of ATG9A in neuronal health and development [91].

ATG9 vesicles are generated from the secretory pathway and navigate through various cellular compartments, including the ER, Golgi, trans-Golgi network, plasma membrane, and endosomal compartments. Using cells deficient for both ATG2A and B to generate a pre-ATG2 compartment that is enriched for ATG9 and MAP1LC3B-II, researchers find that upon autophagy induction, ATG9 vesicles are recruited to the phagophore assembly site (PAS) to serve as seed vesicles to initiate autophagosome biogenesis [81]. ATG9 possesses a lipid scramblase activity, which together with the lipid transferase ATG2 facilitates the exchange of phospholipids between the inner and outer leaflets of membranes [80,196]. This activity is vital for the expansion and maturation of the phagophore, as it helps in lipid mobilization from lipid droplets to phagophore membranes [197].

Much less is known about how ATG9 regulates autophagosome formation in neuronal cells [198]. ATG9-containing vesicles originate from the trans-Golgi network located in the soma and must travel to distal axons where autophagosome biogenesis is prominent. These vesicles are transported anterogradely along axons by kinesin motors, such as UNC-104/KIF1A, which are responsible for moving cargo toward the plus ends of microtubules [199]. Once at the synapse, ATG9-containing vesicles undergo activity-dependent exo-endocytosis, a process that is regulated by proteins such as UNC-13/Munc13 and UNC-26/Synaptojanin [72]. ATG9 couples synaptic exo-endocytosis to autophagy, thus linking synaptic autophagy to the activity state of the neuron. The authors propose that disruptions in endocytosis or autophagy could impair the clearance of damaged synaptic components, ultimately leading to synaptic dysfunction [72].

Very recent evidence using super-resolution microscopy and proteomic analyses reveal that ATG9 vesicles possess a unique protein composition, with low levels of traditional trafficking proteins such as coat proteins, tethering factors, and SNAREs [85,86]. In C. elegans neurons, the protein CLA-1/clarinet, particularly its long isoform (CLA-1 L), plays a crucial role in the active zone, the specialized, electron-dense region on the presynaptic terminal where synaptic vesicles dock, prime, and fuse to release neurotransmitters. CLA-1 regulates the sorting and trafficking of the autophagy protein ATG9 at synapses. This regulation is essential for presynaptic autophagy, helping to maintain synaptic protein homeostasis [78]. CLA-1 facilitates the sorting of ATG9-containing vesicles, ensuring they are properly trafficked to synapses where they can participate in autophagosome formation. This process is particularly important during increased neuronal activity, as it helps manage the demand for enhanced autophagy flux. Additionally, CLA-1 interacts with endocytic scaffolding proteins, linking exocytosis at the active zone with endocytosis at the periactive zone. This coordination is vital for the efficient cycling of synaptic components and the maintenance of synaptic function [78].

Overall, ATG9-containing vesicles are specialized for lipid delivery in autophagy, operating in parallel to, but distinct from, synaptic vesicle cycles. Their distinct composition suggests that ATG9 vesicles do not directly participate in the typical synaptic vesicle cycle but rather function as lipid shuttles distinct from synaptic vesicles. ATG9 vesicles scavenge lipids from intracellular compartments to support autophagosome formation, a process crucial for maintaining synaptic protein homeostasis and supporting neuronal function. The vesicles likely facilitate direct lipid transfer at synaptic sites, compensating for the limited ATG2-mediated lipid transfer seen in non-neuronal cells [198]. This specialized adaptation highlights the critical role of ATG9 in neuronal autophagy and synaptic maintenance. These studies also emphasize the heterogeneity in membrane composition of ATG9-containing vesicles in different cell types and cell compartments.

Microglia, the resident immune cells of the central nervous system, are involved in the surveillance and maintenance of synaptic environments. Recent evidence demonstrates loss of Atg9 in fly glial cells leads to reduced autophagic activity and accumulation of glial protein aggregates, accompanied by progressive dopaminergic neuron loss and locomotion deficits as seen in PD [200]. GAK (cyclin G associated kinase) is a multifunctional protein kinase involved in various cellular processes, including clathrin-mediated endocytosis and cell cycle regulation. In the context of PD, GAK has been identified as a genetic risk factor, with studies suggesting that its dysfunction may contribute to the pathogenesis of the disease. GAK is thought to influence PD by affecting the endocytic pathway within glia and potentially modulating the aggregation and clearance of SNCA/alpha-synuclein, a protein that forms toxic aggregates in the brains of individuals with PD. In the absence of GAK, the number and size of the autophagosomes and autophagosomal precursors increase in adult fly glia and mouse microglia. The Drosophila GAK homolog Aux contributes to PD-like symptoms including dopaminergic neurodegeneration and locomotor function in flies [201].

Autophagy in microglia is essential for synaptic pruning, a process critical for the refinement of neural circuits during development [202]. Impaired autophagy in microglia leads to defective synaptic pruning, resulting in increased dendritic spine density and synaptic dysfunction [203]. Microglia lacking the autophagy-related gene ATG7 exhibit impaired synaptic degradation and an increased number of immature synapses, which are associated with neurodevelopmental disorders like ASD [203]. Conditional knockout of Atg5 in microglia inhibited postnatal neurogenesis in the dentate gyrus of the hippocampus, but not in the subventricular zone of 5×FAD mice, although the link with synaptic pruning was not studied [204]. Other studies suggest that inhibition of autophagy impairs the phagocytic capacity of microglia. For example, Beclin-1-mediated phagocytic dysfunction was found to be associated with impaired recycling of phagocytic receptors [205].

TREM2 (triggering receptor expressed on myeloid cells 2) is an immunomodulatory transmembrane receptor primarily expressed on microglia. Deficiency in TREM2 leads to an energy crisis in microglia, causing mitochondrial damage and triggering autophagy through activation of AMPK and impairment of phosphoinositide 3-kinase/PI3K-AKT-MTOR signaling pathways [206]. The increase in autophagy is thought to be a compensatory mechanism to manage cellular stress and maintain homeostasis. Although the direct link between TREM2-mediated autophagy and synaptic activity is not fully elucidated, the regulation of microglial autophagy by TREM2 can influence synaptic pruning and maintenance. Indeed, TREM2 is essential for microglia-mediated synaptic refinement during the early stages of brain development. The absence of TREM2 results in impaired synapse elimination, which is accompanied by enhanced excitatory neurotransmission and reduced long-range functional connectivity. This impairment is linked to behavioral changes such as repetitive behavior and altered sociability in animal models, which are also associated with neurodevelopmental disorders in humans such as ASD [207].

Autophagy deficiency in microglia is closely linked to increased inflammation, which can exacerbate neurodegenerative processes. A very recent study has identified a novel microglial population in the aging mouse brain that exhibits cytoprotective properties. This microglial subpopulation is characterized by active autophagy and plays a crucial role in protecting against neuroinflammatory damage, such as that observed in multiple sclerosis models. The protective effect of these microglia is autophagy-dependent, as demonstrated by increased neural and glial cell death when the autophagy gene Ulk1 is specifically deleted in microglia [208]. Autophagy impairment in microglia is associated with the activation of inflammasomes, such as the NLRP3 inflammasome, which further amplifies the inflammatory response. For instance, the loss of autophagy-related genes like Atg7 in microglia leads to increased production of inflammatory mediators and exacerbates neuroinflammation in models of neurodegenerative diseases like PD and AD [209,210]. A number of studies have demonstrated a beneficial role of mitophagy in suppressing microglia-mediated neuroinflammation by reducing mitochondrial ROS generation and inhibiting the NLRP3 inflammasome and CGAS-STING1 pathway [210]. Recent studies have demonstrated beneficial effects of mitophagy-inducing drugs such as the postbiotic urolithin A on reducing amyloid-β and MAPT/tau pathologies, and reversing cognitive deficits [211,212]. In aged animals, treatment with urolithin A alleviates age-associated neurological decline, enhancing mitophagy, improving synaptic connectivity, and reducing neuroinflammation [213]. This study highlights the potential of targeting mitophagy to modulate inflammation and improve healthspan during aging, offering insights into the mechanisms underlying age-related neuroinflammatory diseases.

Astrocytes also play a significant role in synaptic regulation, and astrocyte autophagy is implicated in various neurodegenerative and neuroinflammatory conditions. Autophagy in astrocytes is crucial for maintaining cellular homeostasis and supporting neuronal function by clearing damaged proteins and organelles. Furthermore, astrocyte autophagy is implicated in the regulation of synaptic plasticity and neuroinflammation. In major depressive disorder, autophagy-related pathways are significantly inhibited in astrocytes, which may contribute to the pathology of the disorder [214]. Enhancing autophagy in astrocytes has been suggested as a potential therapeutic strategy for alleviating symptoms of major depressive disorder by promoting synaptic health and reducing inflammation [214]. By maintaining synaptic integrity and function, astrocyte autophagy supports cognitive processes and neural circuit stability, underscoring its importance in brain health and disease [215].

HTT, the gene implicated in HD, pathogenically undergoes a polyQ/glutamate expansion in exon 1, resulting in a CAG trinucleotide repeat expansion that disrupts protein structure and canonical function [336]. Wild-type HTT (huntingtin) binds the C terminus of ULK1, an autophagy initiating kinase, and competes with MTORC1 to promote autophagy [337,338]. The polyQ/glutamate tract expansion present in patients with HD disrupts ULK1 interactions leading to impaired autophagic function [222]. Disruption of canonical HTT function due to polyQ expansion has also been reported to impair selective mitophagy [222]. Despite its well-known role in aggregation, variation in polyQ length may confer neuronal protection via changes in transcriptional programming; these authors propose that changes in polyQ expansion size potentially confer neuronal and evolutionary fitness [339].

ATG9 is known to localize to sites adjacent to phagophore formation in the early stages of autophagy and is believed to promote expansion of the autophagosome among other roles [196,340,341]. ATG9A plays a critical role in the mobilization of lipid droplets to the autophagosome and mitochondria, a feed forward signal in autophagy processing [197]. Upstream of its lipid scramblase activity, ATG9A is also a component of and plays a role in the formation and maturation of the ULK1 complex. Recent X-ray crystallography studies have defined the ternary structure of the ATG9S carboxyl-terminal tail bound to ATG13-ATG101 in the ULK1 complex [223]. Disruption of this interaction impairs downstream mitophagy activity [223]. The role(s) of ATG9 in the nervous system are emerging as discussed above, yet still understudied [198].

PIK3R4/VPS15 is a core component of the class III phosphatidylinositol 3-kinase (PtdIns3K) complex required for phosphatidylinositol-3-phosphate (PtdIns3P) generation in the early stages of autophagosome formation [220]. Loss of function mutations associated with neurodevelopmental disorders lead to defects in the assembly of the PtdIns3K complex and blunt subsequent generation of PtdIns3P, inhibiting autophagy and leading to the accumulation of autophagic substrates in mice and humans. In humans, mutant PIK3R4/VPS15 is associated with epilepsy and atrophy in the cortical region of the brain [224]. When ULK1-mediated phosphorylation of PIK3R4/VPS15, a pro-autophagy signal, is disrupted by targeted mutagenesis experimentally, unphosphorylated VPS15 leads to inhibition of autophagy [342]. However, whether disease-linked mutations in VPS15 impair autophagy signaling due to non-canonical mechanisms or due to a lack of phosphorylation by ULK1 has not been shown in the context of disease pathogenesis.

EPM2A/Laforin and NHLRC1/malin/EPM2B form a functional ubiquitin ligase complex that polyubiquitinates key components of the autophagic class III phosphoinositide 3-kinase complex (e.g., BECN1, PIK3C3/VPS34) to enhance PI3KC3 activity and facilitate autophagy initiation [225]. Mutations in the EPM2A gene, a glucan phosphatase, are linked to Lafora disease and disrupt autophagy initiation. Loss of functional EPM2A/laforin reduces PtdIns3P levels and results in defective PI3KC3 complex stabilization and impaired initiation of autophagy [225]. EPM2A/Laforin deficiency has also been implicated in autophagy dysfunction through an MTOR-dependent pathway, exacerbating the accumulation of autophagic substrates and cellular stress [226].

WIPI (WD repeat domain, phosphoinositide interacting) proteins play an essential role in the early stages of autophagosome maturation. Mutations in WDR45/WIPI4 are associated with β-propeller protein associated neurodegeneration (BPAN) [343]. WDR45 plays a role in neural development and is expressed in excitatory synapses of mouse brains. Acute knockdown of WDR45 causes dendritic and synaptic abnormalities during corticogenesis which is rescued when the gene is re-introduced [343]. Loss of WDR45 impairs autophagy and BPAN-associated mutations alter autophagy activity, which is either increased or decreased dependent on the specific mutant expressed [227]. In the context of neurodegeneration, it was recently shown in cell culture and zebrafish models that loss of WDR45 induces ferroptosis independent of autophagy. Without WDR45, ATG2A localizes to ER-mitochondrial contact sites to enhance mitochondrial import of phosphatidylserine, a lipid prone to peroxidation, allowing for ferroptotic cell death [344].

Several mutations in the ATG class of proteins alter transcriptional signaling or protein function and affect autophagosome formation dynamics. In one study, a heterozygous variant of ATG5 in a single female PD patient was identified and demonstrated to enhance the transcriptional activity of the ATG5 promoter [228]. The enhanced transcriptional activity may act to alter autophagic signaling by enhancing aberrant (in this context) formation of the ATG12–ATG5-ATG16L1 complex. Variants of ATG7 in neurodegenerative diseases reduce autophagic activity [229]. Childhood ataxia or cerebral palsy-associated mutations in ATG5 ablate conjugation to ATG12 and/or protein expression, impairing autophagosome formation and autophagic flux [345,346]. Polymorphisms in the ATG7 promoter region reduce transcription and are associated with earlier onset of HD [347,348] and sporadic PD [349]. These variants elicit reduced formation of the ATG12–ATG5-ATG16L1 complex, which is canonically accomplished by ATG7 activity [229]. PD-associated mutations upstream of exon 1 modulate ATG12 transcription, likely due to structural genomic differences that alter promoter interactions with different transcription factors [230]. Although inferences could be made regarding how these mutations might elicit transcriptional changes to alter autophagy, additional in vitro and in vivo evidence are needed to clarify the roles that these mutations may play in regulating autophagy in the context of disease pathogenesis.

While a myriad of previous work has characterized the many protein pathologies in neurodegenerative disorders that are either mechanistically linked to or correlated with autophagy dysfunction, contemporary studies seek to assess autophagy genes as biomarkers in neurodegeneration. Perhaps the greatest challenge in the evaluation of autophagic activity in humans is that autophagy is a dynamic process whereas measurements from patient samples only provide static "snapshots" of the autophagosome markers that may or may not correlate with activity or flux. Measurement of MAP1LC3-II (lipidated MAP1LC3-I) to housekeeping protein ratios are a classic method of determining if cells exhibit increased autophagosome levels, but such measurements taken alone do not inform about upstream processes affecting autophagy initiation or downstream completion of lysosomal degradation. The snapshot is generally informative regarding autophagic dysregulation, but current methods do not provide for the determination of autophagic flux (e.g., elevations in autophagosome content could result from robust autophagy initiation and/or from impaired intracellular trafficking and lysosomal dysregulation). Ideally, evaluation of autophagy would include clear indicators of autophagosome formation and degradation, dependent upon the autophagy machinery, and functional readouts relating autophagy parameters to cell health; unfortunately, this approach currently lacks feasibility, would be very expensive, and requires standardized definitions or benchmarks of autophagy processes in specific regions of the brain (e.g., basal and stressed flux ratios, kinetics of initiation under defined stimulation conditions, lysosome activity, etc.) in healthy individuals, with which to compare against patient samples. Even in more controlled, inbred animal model systems, the development of accurate and dynamic biomarkers of autophagy are needed to better understand the molecular pathology of autophagy dysregulation.

Given the importance of PTMs in governing the activity or functional status of many autophagy and mitophagy proteins, valuable information may be inferred from their study. Although still representing static readouts, understanding PTMs and multiplexed assessment of many autophagy genes would at least allow discrimination of activation states for such proteins. Ultimately, machine learning, artificial intelligence or other systems-based methods may be needed to develop an improved understanding of how flux states are disrupted in individuals and populations affected by neurodegenerative disease. Here, we highlight potential future prognostic biomarkers seeking to define disease severity and the impact on autophagy/mitophagy among patients.

In addition to the well-known lipidation of species required for intracellular membrane incorporation, MAP1LC3 and other Atg8-family proteins are also modified by phosphorylation and other PTMs that further regulate their interactions and functions in governing autophagy processes [484,485]. Similarly, the various autophagy receptors are complexly modified during autophagy [486]. Understanding the PTM code of protein groups (e.g., autophagy receptors) paired with technical advances to monitor PTM changes of protein involved in autophagy may allow for real-time evaluation of autophagy initiation, progression, and execution – given that basal function of such targets is well-defined. Also for consideration are the individual protein domains modified by PTMs. Both the type of PTMs and the unique sites they modify can exert distinctly different functional effects on the protein in question.

Biochemical and functional analyses in recent years revealed that SQSTM1 can be phosphorylated in its oligomerization, MAP1LC3-interacting, or ubiquitin binding domains, which consequently direct it to be more or less engaged in aggrephagy, mitophagy, or secretory autophagy [38]. The SCOC protein was not previously established as an autophagy receptor, but one study identified a LIR-related sequence, which conferred SCOC autophagy receptor/adaptor action following phosphorylation of serine residues adjacent to the LIR [487]. This indicates that potentially many more LIR-related proteins could behave as autophagy receptors following some PTM that enhances its affinity for MAP1LC3 or similar Atg8-family proteins. PTMs of MAP1LC3 (including ubiquitination, lipidation, acetylation, and phosphorylation) can throttle its affinity for autophagy receptors, resulting in various effects on autophagy [484,485]. In theory, this could mean that uncovering specific PTMs of MAP1LC3 combined with PTMs regulating autophagy receptors may reveal the snapshot of current efforts in the autophagy landscape (e.g., more aggrephagy, mitophagy, or generalized autophagy). Combined evaluation of phosphorylation states for multiple key proteins could reveal the prevailing autophagy signaling mechanism in a given sample.

Additionally, upstream analyses of kinase signaling/activity would provide useful information on whether the background cellular context is permissive or inhibitory for autophagy. TBK1 is a kinase that phosphorylates various autophagy receptors and MAP1LC3-II, providing a feed forward signal for autophagy. However, different sites of MAP1LC3 and other GABARAP phosphorylation can differentially regulate lipidation or other biochemical features which impact the progression of autophagy at different stages [488], providing yet another layer of complexity to be decoded with regard to site specific combinatorial effects. One study newly found that phosphorylated TBK1 accumulates in protein aggregates in an AAV-C9orf72(G₄G₂)₁₄₉ mouse model [489]. Introduction of a reduced activity TBK1 mutant in this mouse model exacerbated phenotypes including an increase in TARDBP/TDP-43 pathology and accumulation of irregular endosomes [489]. Utilization of accessible biomaterials or alternative methods for in vivo evaluation of pTBK1 in the brain (e.g., PET-CT using radiolabeled antibodies) may be considered.

As another example, PINK1-mediated phosphorylation of RHOT/MIRO at different sites have opposite effects on RHOT/MIRO stability; combinatorial studies show that PINK1-mediated stabilization of RHOT/MIRO is the dominant effect [163]. One review also proposes monitoring the oxidation state of the antioxidant scavenging PARK7/DJ-1 in relation to cellular pathology with respect to ROS [264]. Working toward more thorough descriptions of various PTMs, the sites they modify, and how different types of PTMs work in concert on the same proteins, would further enhance the ability of an individual snapshot of biomarkers to reflect dynamic autophagy processes.

As discussed above, acute loss of mitochondrial membrane potential activates the PINK1-PRKN mitophagy pathway, in which OMM proteins become labeled with poly-ubiquitin chains including p-S65-Ub by the coordinated activities of PINK1 and PRKN. While measurement of non-modified PINK1 and PRKN may be useful in evaluating general protein stability in the contexts of disease-causing mutations, understanding protein activation status would be more informative to unraveling the functional capability. Measuring homeostatic PINK1 levels is a challenge due to the continual turnover of this ubiquitin kinase in absence of mitochondrial uncoupling. Yet, PRKN protein levels may be used as an indirect indicator of pathway activity as this ubiquitin ligase seems to be "consumed" with the degraded mitochondria; a complete loss of PINK1 stabilizes PRKN protein in its inactive form [365,490,491]. PINK1 and PRKN phosphorylation, with respect to PD-linked mutations, have been assessed in vitro, and efforts of developing quantitative and highly specific assays to measure these PTMs in PINK1 and PRKN are ongoing and would provide useful insights into the effect of mutations on activation status and signaling in vivo.

The mitophagy signal, p-S65-Ub, is a molecule specifically generated by the activity of PINK1 and represents < 1% of endogenous ubiquitin. The levels of p-S65-Ub increase during the normal aging process, and are higher in patient brains affected by AD [492], Dementia with Lewy Bodies [493] and PD [360], but not multiple system atrophy [494]. Increased levels of p-S65-Ub with aging are indicative of greater mitochondrial dysfunction requiring surveillance by the mitochondrial quality control system. Decreased p-S65-Ub may indicate low activation of this pathway, either due to low requirement for turnover (due to limited stress/damage) or mutation-associated disruption of p-S65-Ub production or mitochondrial labeling, or reflect engagement of PINK1-independent mitophagy pathways. Increased p-S65-Ub in the context of neurodegenerative disorders (compared to healthy individuals) may signal either impaired turnover or increased active turnover to compensate for pathology-associated mitochondrial stressors; therefore, understanding an individual's pathologies (including other mutations in autophagy/mitophagy genes) is crucial for the accurate interpretation of p-S65-Ub levels. Recent and ongoing efforts have allowed for greater precision and specificity in measuring p-S65-Ub [495,496]. Levels of p-S65-Ub have been exploited for functional screening in cells and more recently as a quantitative trait in an autopsy brain genome-wide association study, resulting in the identification of disease risk and potential resilience factors [457,497,498]. While this biomarker is not informative concerning other mitophagy pathways (Table 1), levels of p-S65-Ub may also be used as a biomarker to track increased PINK1 activity in living systems in response to small molecules that inhibit its degradation [174] or promote its activity [499].

These additional modalities should be considered as they each provide unique insight into the activation state and status of proteins, allowing for more precise prognostic evaluation at the molecular level regulating autophagy activation and progression. Working toward the building of standardized panels that define PTMs in "key" autophagy genes via highly specific and sufficiently sensitive assays, and using them to evaluate large well-defined and well-controlled clinical cohorts would improve the precision by which autophagy dynamics can be described. This would of course require the generation and standardization of normal autophagy dynamics "benchmarks" in healthy men and women of different genetic backgrounds that could be compared to diseased samples. Such studies will improve our general understanding of how autophagy is compromised during various stages of neurodegeneration in a disease-specific manner and serve as screening tools to determine which treatments (i.e. what gene or part of the pathway to target) would best serve individuals suffering from neurological diseases.

MTOR (mechanistic target of rapamycin kinase) is a component of the MTORC1 and MTORC2 complexes that negatively regulate autophagy by phosphorylation of ULK1 at its S757 inhibitory residue; MTOR is hyperactive in AD which inhibits autophagic clearance of pathologically associated protein aggregates [500]. Efforts to target overactive MTOR signaling have long been in development as a means of enhancing autophagic clearance of protein aggregates or damaged intracellular compartments. The evidence supporting rapamycin as a potential treatment for AD or other neurodegenerative disorders is mostly pre-clinical. A phase I clinical trial was successful in establishing safety among patients with AD [501], and there are currently two phase II studies still in the recruitment stage [502]. MTOR is also hyperactive in PD, HD, and ALS, suggesting that these populations may also benefit from rapamycin treatment [500].

Metformin [503] and trehalose [504] are thought to activate autophagy through activation of AMPK, not requiring inhibition of MTOR, but more recent evidence has also shown that trehalose also induces autophagy by activation of TFEB [505,506]. A 1–2 year treatment with metformin in normal and AD mice demonstrated cognitive improvement in mice younger than 16 months old, whereas longer treatments caused cognitive decline in healthy mice and mice with AD; the authors cautioned that care should be taken when considering metformin treatment in patients with AD and highlighted conflicting studies as to the safety and efficacy of metformin [507]. Other caveats regarding the use of metformin come from a large meta-analysis on its utility in preventing dementia in patients with or without diabetes (metformin is widely used as a treatment for diabetes), where the results suggest it may not be effective in the prevention of dementia in patients without diabetes [508]. Despite this, Metformin has been used to alleviate autophagic inhibition in animal models of PD and HD as well as in patients with AD, PD, and HD. In general animal models show improvement, but efficacy and side-effects in patients and animal models (such as reduced vitamin B12 levels) should be further explored to refine treatment with Metformin, or to better identify sub-populations that would benefit most from treatment [503].

Trehalose has been assessed in several animal models of neurodegenerative disease (PD, AD, HD, ALS) and found to reduce protein aggregates and rescue autophagy function. In a seminal article published in 2013, Trehalose prolonged lifespan, increased autophagy in motoneurons, and decreased aggregation of mutant SOD1 protein in an ALS mouse model, leading to another decade of research on MTOR-independent autophagy activation. The molecular characteristics, mechanism of action, and current and future disease indications for Trehalose therapies has been extensively reviewed [509]. Pre-clinical data demonstrates neuroprotection with limited pleiotropic affects in various disease models. Early phase clinical trials using Trehalose have demonstrated safety and improved cognitive and motor neuron functions (e.g., swallowing) [509]. A trial using trehalose to treat ALS demonstrated efficacy among individuals not taking RELYVRIO (another medication used to treat ALS that was recently discontinued after a failed phase III clinical trial; [https://clinicaltrials.gov/study/NCT05136885?tab=results#outcome-measures↗, last accessed 22 Aug 2025]). Taken together, trehalose seems to have promising therapeutic potential for neurodegenerative diseases in which autophagic function is disrupted.

Acting at the terminal step in autophagy, lysosomes regulate autophagic turnover and are dysregulated in several neurodegenerative diseases. Targeting lysosomal status is particularly important given the known detrimental consequences of driving autophagy initiation in cells with lysosomal impairment [192]. Lysosomes regulate autophagy via inactivation of MTOR during cellular nutrient deficiency, and lysosomal deacidification impairs the activity of pH sensitive lysosomal enzymes [456]. Pharmacological approaches to enhancing lysosomal acidification rescue autophagic degradation and prevent inflammation in mouse models of lysosomal storage disease [510,511]. Several drugs restore the acidic pH of the lysosome via modulation of the MCOLN/TRPML ion channel or V-ATPase activity [512], and delivery of nanoparticles that enhance lysosomal acidification are promising and under early development [513]. As nanoparticle medicines are highly heterogeneous and represent an evolving area in biology, these therapies may require additional regulatory considerations compared to "simple drugs" [514].

Inhibitors of LRRK2 activity (either pharmacological kinase inhibitors or anti-sense oligonucleotides that decrease LRRK2 expression) demonstrate promising results in pre-clinical and early phase clinical studies for PD. One early phase II study seeks to enhance lysosome function via activation of GBA and another phase II study in progress seeks to restore GBA function using adeno-associated-virus (AAV) gene therapy to replace mutant GBA in patients with PD [400]. Another gene therapy targeting lysosomes restores the nominal function of PGRN via delivery of functional PGRN, referred to as ALPHA-0602. This pre-clinical gene therapy is reported to restore PGRN levels in vitro and in vivo, protects mice against TARDBP/TDP-43 pathology, and reduces motor neuron toxicity in an ALS model [515]. ALPHA-0602 has an orphan drug designation in the United States (https://www.accessdata.fda.gov/scripts/opdlisting/oopd/detailedIndex.cfm?cfgridkey=725319↗, last accessed 22 Aug 2025), but currently there are no clinical studies recruiting for ALPHA-0602.

Other potential avenues for gene therapies include DJ-1, NDP52, PINK1, and PRKN. Introduction of wild-type DJ-1 seems to be neuroprotective in rats against neuronal death in a model of 6-hydroxydopamine induced PD, congruent with similar findings that pharmacologically targeting DJ-1 is protective in these contexts [516]. The discovery of a protective and more active NDP52 variant that interacts with autophagic machinery more readily compared to wild-type NDP52 implicates NDP52 as a novel gene therapy target for AD [371]. Another potential target for gene therapy is the mitophagy associated kinase PINK1, where in vitro work demonstrates enhanced production of p-S65-Ub by a G411A mutant associated with increased mitochondrial turnover [368]. Although this PINK1 variant has not been identified in a patient, the neuroprotective effects it confers may be able to limit the neuropathology or may provide resilience. Animal models of AD and HD show enhanced neuroprotection with AAV-PINK1 overexpression or germline editing [172,173,517]. Similarly, models of PD and AD show that overexpression of PRKN (by means of AAV, lentivirus, germline editing, or introduction of membrane permeable PRKN) reduces cellular pathology and protects against disease progression [518–522]. Considering the many known autophagy and autophagy-related genes with causal mutations in disease progression that perturb normal function, and the major advances in gene editing technology over the last 20 years, additional targets of autophagy may be considered. The benefit of targeting autophagy as a mechanism, especially with causal genes implicated in multiple disorders (e.g., SNX14), is that one gene therapy could show applicability across multiple diseases, potentially allowing for more rapid translation of such treatments.

Mitophagy can also be enhanced pharmacologically, and several strategies target PINK1, PRKN, USP30, and SQSTM1– a few of which have recently entered clinical trials [523]. Urolithin A is a naturally occurring metabolite that stimulates mitophagy by promoting PINK1 and PRKN activity by a mechanism that is not entirely understood [524]. Safety was demonstrated in healthy individuals after in vivo testing showed prolonged lifespan in C. elegans and improved muscle function in rodents [525,526]. Urolithin A has also been shown to improve multiple cognitive measures and to reduce neuropathology in mouse models of AD, physiological aging and a model of macular degeneration [527,528]. One of the earliest small molecules targeting PINK1, kinetin triphosphate, is an ATP analog that was initially shown to increase both a mutant and wild-type PINK1 activity in vitro [499]; although direct binding of kinetin triphosphate to PINK1 was recently questioned [529]. Inhibiting the rapid degradation of PINK1 is another strategy to elevate the bioavailability of this neuroprotective kinase. The small molecule BC1464 interferes with the interaction of PINK1 with the ubiquitin ligase FBXO7, elevating PINK1 expression and p-S65-Ub levels, and protecting against cell death and dendritic atrophy in chemical and genetic models of PD [174]. Another small molecule, BIO-2007817, promotes PRKN activity via direct interaction and can rescue the activity of some disease mutants [530,531]. Other small molecule activators of PRKN were identified in a high-content screen and found to exert their action via inhibition of USP30 [532]. Deubiquitinases, such as USP30, inhibit PINK1- and PRKN-mediated mitophagy by the deconjugation of ubiquitin chains on OMM species, thus limiting the ability of ubiquitin-dependent mitophagy adaptors to tether OMM proteins to the forming autophagosome. Inhibition of USP30 enhances mitophagy in vitro and can rescue p-S65-Ub levels in fibroblasts with pathogenic PRKN mutations [533,534]. Knockout or inhibition of USP30 protects against loss of dopaminergic neurons in a mouse model of PD [535]. Similarly, inhibition of USP14 (another deubiquitinase) can rescue mitophagy activity in vitro, but apparently in a PRKN-independent fashion [39]. The autophagy receptor SQSTM1 has also been targeted to enhance mitophagy in vitro [536]. One caveat to consider is that each of these proteins have additional functions beyond autophagy, and it may be difficult to ascribe beneficial or detrimental impacts on one particular pathway.

Last, it would be negligent to disregard the impact of lifestyle choices on disease progression. Caloric restriction (CR) has been long recognized as major factor that influences brain aging and the cognitive decline associated with neurodegeneration [537]. By nature of the low nutrient conditions under which autophagy is activated, restricting food intake stimulates autophagy [538]. A thorough review summarizes the effects of caloric restriction on various animal models of neurodegeneration and reports that CR regulates neuroinflammation in neurodegenerative diseases [539]. Another article discusses how CR reduces symptomology and disease progression in various disorders [540]. A recent report demonstrates that 40% CR led to a 9 month ( > 36%) increase in lifespan compared to mice that were fed ad libitum [541]. One caveat to consider is the degree of CR that may be necessary to induce autophagy in the brain. In mice, 48 h of complete starvation does not induce brain autophagy [542], most likely because responses in other peripheral tissues act to support the nutritive status of the brain. While neurons in culture do respond to nutrient deprivation by upregulating autophagy, these responses are more robust in male neurons, with female neurons showing primarily lipid mobilization [543]. Thus, beneficial impacts of CR may involve multiple mechanisms.

Another factor known to prolong lifespan in humans is exercise. Unsurprisingly, exercise stimulates autophagy and is thought to be neuroprotective, both for healthy aging and in the context of neurodegenerative disease [544,545]. Exercise may also be beneficial for reducing age-related cellular senescence [546]. Again, most of these studies are correlative, showing upregulation of autophagy coincident with the beneficial effect being studied. In aged mice, exercise-related improvements in mitochondrial function were not linked to increased autophagy, but to changes in mitochondrial fission [547]. Nevertheless, as CR and exercise are generally accessible despite factors such as socioeconomic status, which may limit patient access to certain therapies, they should be considered and further studied as a means of ameliorating or delaying symptom onset in susceptible families and individuals.