Download File

Mechanisms of Dynamic Recruitment of the ESCRT Pathway in Axons

Veronica Kate Birdsall

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy under the Executive Committee of the Graduate School of Arts and Sciences

COLUMBIA UNIVERSITY

2020

Abstract

Mechanisms of Dynamic Recruitment of the ESCRT Pathway in Axons

Veronica Kate Birdsall

Clearance of molecularly damaged and misfolded synaptic vesicle (SV) proteins is vital for the maintenance of healthy, functional synapses. However, this process poses significant trafficking challenges for neurons, as the majority of degradative organelles and machinery are localized in the somatodendritic compartment, far from SV pools in presynaptic terminals. Our previous work showed that SV protein degradation is mediated by the endosomal sorting complex required for transport (ESCRT) pathway in an activity-dependent manner. Moreover, we found that neuronal activity increased ESCRT protein recruitment to axons and SV pools, suggesting a novel mechanism for regulating the trafficking of this critical degradative machinery, whose localization and transport in neurons has been unexplored. Here, we characterize the axonal transport of ESCRT-0 proteins Hrs and STAM1, the first components of the ESCRT pathway, which are critical for initiating SV protein degradation. We find that Hrs- and STAM1-positive transport vesicles exhibit increased anterograde and bidirectional motility in response to neuronal activity, as well as frequent contact with SV pools. ESCRT-0 vesicles typically colocalize with early endosome marker Rab5, but their transport dynamics do not mirror those of the total Rab5 vesicle pool. Moreover, other ESCRT pathway components and effectors do not show activity-dependent changes to motility, indicating that neuronal firing specifically regulates the motility of the ESCRT-0+ subset of Rab5+ structures in axons. Finally, we identify kinesin-3 motor protein KIF13A as essential for the activity-dependent transport of

ESCRT-0 vesicles as well as the degradation of SV membrane proteins. Altogether, these studies demonstrate a novel activity-dependent mechanism for mobilizing the axonal transport of a

newly characterized endosomal subtype carrying ESCRT machinery. This activity-induced transport is necessary for ESCRT-mediated degradation of synaptic vesicle proteins.

Table of Contents

LIST OF FIGURES ...... iii ACKNOWLEDGMENTS ...... iv CHAPTER 1: GENERAL INTRODUCTION ...... 1 SYNAPTIC HEALTH AND MAINTENANCE ...... 1 Synaptic Breakdown and Neurodegeneration ...... 1 Synaptic Protein Turnover ...... 1 SYNAPTIC AUTOPHAGY...... 2 Autophagy in Synapse Development ...... 2 Autophagic Cargo at the Synapse ...... 4 Outstanding Questions ...... 10 ESCRT PATHWAY ...... 12 ESCRT Pathway Components and Mechanics ...... 12 Regulation of Membrane Protein Degradation ...... 16 ESCRT PATHWAY IN NEURONS...... 18 Degradation of Synaptic Vesicle Proteins ...... 18 Role in Neurodegeneration ...... 19 ENDOSOMAL TRANSPORT DYNAMICS ...... 20 Mechanics of Axonal Transport ...... 20 Autophagosome Transport ...... 21 Endosomal Transport ...... 21 CHAPTER 2: MATERIALS AND METHODS...... 23 CELL CULTURE ...... 23 Neurons...... 23 Cell Lines ...... 23 MICROFLUIDIC CHAMBERS ...... 23 DNA CONSTRUCTS ...... 25 PHARMACOLOGICAL TREATMENTS ...... 26 EVALUATION OF NEURONAL HEALTH ...... 26 LENTIVIRAL TRANSDUCTION AND NEURONAL TRANSFECTION ...... 27 ELECTRON MICROGRAPH ANALYSIS OF ENDOSOMAL STRUCTURES ...... 28 CORRELATIVE LIGHT AND ELECTRON MICROSCOPY ...... 28 IMMUNOFLUORESCENCE MICROSCOPY ...... 30 IN VITRO KIF13A/KIF13B TRANSPORT ASSAY ...... 31 LIVE IMAGING ...... 31 LIVE IMAGING ANALYSIS ...... 32 IMMUNOPRECIPITATION AND WESTERN BLOT ...... 32 STATISTICAL ANALYSES ...... 33 CHAPTER 3: ESCRT-0 PROTEINS HRS AND STAM ARE BIDIRECTIONALLY MOTILE IN AXONS, AND UNIQUELY REGULATED BY NEURONAL FIRING ...... 34 ABSTRACT ...... 34 RESULTS...... 35 Hrs and STAM are bidirectionally motile at baseline, and exhibit increased motility in response to neuronal firing ...... 35 ESCRT-I component TSG101 and small GTPase Rab35 do not show activity-dependent motility changes ..... 39 SUMMARY ...... 44

CHAPTER 4: ACTIVITY-RESPONSIVE HRS+ VESICLES ARE A SUBSET OF TOTAL RAB5+ VESICLES ...... 46

RATIONALE ...... 46 RESULTS...... 48 Rab5 shows a modest activity-dependent increase in axonal motility ...... 48 Hrs+ vesicles are a subset of Rab5+ vesicles ...... 50 Hrs targeting to vesicle membranes depends on the FYVE domain/PI3P interaction and occurs in the soma. 53 SUMMARY ...... 54 CHAPTER 5: HRS-POSITIVE VESICLES INTERACT WITH SV POOLS ...... 56 RATIONALE ...... 56 RESULTS...... 58 Motile Hrs puncta colocalize with stationary Synapsin puncta in the axon ...... 58 CLEM reveals that axonal Hrs-positive vesicles are degradative structures ...... 58 Endosomal structures are more prevalent at the synapse in high-activity conditions ...... 60 SUMMARY ...... 62 CHAPTER 6: KIF13A IS RESPONSIBLE FOR ACTIVITY-DEPENDENT ANTEROGRADE MOVEMENT OF HRS IN THE AXON, AND NECESSARY FOR THE TURNOVER OF ESCRT-DEGRADED SV PROTEINS ...... 63 RATIONALE ...... 63 RESULTS...... 65 KIF13A, but not KIF13B, interacts with Hrs+ vesicles ...... 65 Knockdown of KIF13A ablates activity-dependent increase in ESCRT-0 motility and slows degradation of SV proteins ...... 68 SUMMARY ...... 70 CHAPTER 7: DISCUSSION AND FUTURE DIRECTIONS ...... 74 AUTOPHAGY AND THE ESCRT PATHWAY ...... 76 SPECIFICITY OF ACTIVITY-DEPENDENT REGULATION TO ESCRT-0+ VESICLES ...... 77 IDENTITY OF ESCRT-0+ VESICLES...... 79 HRS ENDOSOMAL TARGETING ...... 81 ACTIVITY-DEPENDENT TRANSPORT OF ESCRT-0+ VESICLES ...... 82 CONCLUSION ...... 85 REFERENCES ...... 86

List of Figures

Chapter 1 Figure 1: Proposed regulators and cargo of synaptic autophagy………………………………………..9 Figure 2: Diverse membrane structures observed at Bassoon/Piccolo DKD synapses………………..11 Figure 3: The ESCRT Pathway…………………………………………………………………………..13 Figure 4: Relevant domains of ESCRT-0 proteins Hrs and STAM……………………………………14

Chapter 2 Figure 5: Schematic diagram of microfluidic chambers and lentiviral transduction…………………25

Chapter 3 Figure 6: Characterization of cell health, lentiviral transduction efficiency, and neuronal activity effects in microfluidic chambers…………………………………………………………………………..36 Figure 7: Neuronal activity stimulates the transport of ESCRT-0+ vesicles in axons………………...38 Figure 8: ESCRT-I protein TSG101 shows little change in motility in response to neuronal activity.41 Figure 9: Neuronal activity does not stimulate Rab35 motility in axons…………………………….....43

Chapter 4 Figure 10: Rab5+ vesicles show a modest increase in motility in response to heightened neuronal activity……………………………………………………………………………………………………...47 Figure 11: Hrs is present on a subset of Rab5+ vesicles in axons……………………………………....49 Figure 12: Hrs 2xFYVE domain does not recapitulate the motile behavior of full-length Hrs………51 Figure 13: Hrs vesicle association requires FYVE domain/PI3P interactions and occurs in the somatodendritic compartment……………………………………………………………………………52

Chapter 5 Figure 14: Motile Hrs interacts with presynaptic sites in a ubiquitin-dependent mechanism……….57 Figure 15: Correlative light electron microscopy (CLEM) reveals that Hrs+ vesicles can be degradative structures…………………………………………………………………………………….59 Figure 16: Neuronal activity increases the number of presynaptic endosomes……………………….61

Chapter 6 Figure 17: KIF13A and KIF13B mediate transport of Rab5+ vesicles in N2a cells…………………..64 Figure 18: KIF13A, but not KIF13B, mediates transport of Hrs+ vesicles…………………………....67 Figure 19: Depolarization increases the association between KIF13A and Hrs………………………68 Figure 20: Knockdown of KIF13A by sh13A expression in neurons…………………………………..70 Figure 21: Knockdown of KIF13A slows SV protein degradation……………………………………..71 Figure 22: Knockdown of KIF13A inhibits activity-dependent transport of ESCRT-0 proteins…....72 Figure 23: Model of how activity stimulates the anterograde transport of ESCRT-0+ vesicles……..75

iii

Acknowledgments

A profound thank you to past and present members of the Waites and Hengst labs, especially

Clarissa Waites, Patty Sheehan, Mei Zhu, João Luís Vaz Silva, Qi Jin, Tori Zhuravleva, and Lisa

Randolph for providing me with encouragement, scientific expertise, and friendship these past several years. You have taught me everything I know about being a scientist, and I could not have done it without you.

I would also like to thank my friends and family, particularly my parents and Kevin, for their endless support and love during my PhD, and the entirety of my existence.

Finally, I would like to thank my thesis committee for your invaluable guidance and unquestionable patience, both in conducting me through my PhD, and in reading this monstrosity. Thank you for all of your help.

Chapter 1: General Introduction

Synaptic Health and Maintenance

Synaptic Breakdown and Neurodegeneration

Neurons are cells with a fundamentally communal function. They exist, in essence, to interact with each other, and the resulting interneuronal communications form the basis of mammalian life. Synapses, the conduits through which neurons interact, are therefore essential to neuronal function, and synaptic health is fundamental to the health of a brain as a whole.

Studies of neurodegenerative disease have shown that presynaptic dysfunction often precedes cell death (Gorman 2008; Chi, Chang, and Sang 2018). Moreover, ‘synaptopathy’ and synaptic dysfunction are endemic to all neurological diseases regardless of etiology, symptomology, mortality, or age of onset (Taoufik et al. 2020). It is therefore clear that long-term synaptic homeostasis, and the cellular pathways that maintain it, are vital to the prevention of neurological disease.

Synaptic Protein Turnover

Synaptic vesicles (SVs) are the fundamental units of neurotransmitter storage and release, and as such their efficient recycling is essential for neuronal communication. These processes depend upon maintaining the proper complement of functional SV membrane proteins, as the inability to clear nonfunctional or misfolded SV proteins can precipitate synaptic dysfunction and neurodegeneration (Bezprozvanny and Hiesinger 2013; Esposito, Ana Clara, and Verstreken

2012; Hall et al. 2017). SV proteins begin to lose functionality and are taken out of the active pool as they age (Truckenbrodt et al. 2018), presumably due to an accumulation of molecular

damage that occurs over the lifetime of a protein. Previous work has shown that cytosolic synaptic proteins are degraded via the ubiquitin-proteasome system (UPS) and macroautophagy

(Cohen and Ziv 2017), while membrane-associated synaptic proteins, including synaptic vesicle proteins, are degraded under basal conditions by the endolysosomal system (Sheehan et al. 2016;

Uytterhoeven et al. 2011, 2017). Both the macroautophagy and endolysosomal degradative pathways are thus vital to synaptic health.

Synaptic Autophagy

*Note: A version of following section (every sub-heading under ‘Synaptic Autophagy’) has been published in Neuroscience Letters, and the full work can be found here: https://doi.org/10.1016/j.neulet.2018.05.033

Macroautophagy, hereafter referred to as ‘autophagy’, is one of the primary cellular degradative pathways. Although it has only recently become a focus of research in synaptic biology, emerging studies indicate that autophagy has essential functions at the synapse throughout an organism’s lifetime.

Autophagy in Synapse Development

Studies in genetic organisms indicate the importance of autophagy and autophagic machinery in early synaptic development. For instance, impairment of autophagy via the knockdown of ATG1, -2, -6, or -18 was found to reduce the number and size of presynaptic terminals at the neuromuscular junction (NMJ) in Drosophila melanogaster (Shen and Ganetzky

2009). More recently, Stavoe and Colon-Ramos showed that autophagy is required cell- autonomously for presynapse formation in AIY neurons of C. elegans (Stavoe et al. 2016). They

found that autophagosome biogenesis occurred at developing AIY presynaptic boutons. They then observed that 18 distinct components from every stage of the autophagy pathway – initiation (UNC-15, ATG-13), nucleation (ATG-9), elongation (LGG-1, LGG-3, ATG-3, ATG-

4), and retrieval (ATG-2, ATG-9, ATG-18)– were required for the clustering of synaptic vesicle

(SV) proteins at nascent presynaptic sites during AIY neuron development (Stavoe et al. 2016).

However, these components were not required for SV clustering in other C. elegans neuron types, suggesting a cell-specific role for autophagic machinery in presynaptic development.

These findings highlight the importance of autophagy for synaptogenesis, and the need for future studies to clarify its roles in different cell types and organisms.

In the mammalian brain, synapse formation is followed by a period of synapse refinement, and autophagy has been shown to play an important role in this process. Tang and colleagues found that autophagy is required for the developmental pruning of dendritic spines, and that this pruning is disrupted in individuals with autism spectrum disorder (ASD) (Tang et al.

2014). Intriguingly, the authors found a statistically significant increase in phosphorylated mTOR (p-mTOR), the activated version of mTOR, in the brains of ASD patients compared to age-matched controls. Since the activation of mTOR inhibits autophagy, an increase in mTOR activity typically indicates a reduction of autophagy (Tang et al. 2014). Indeed, ASD patients exhibited significantly lower levels of LC3-II, a marker for autophagosomes, throughout childhood and adolescence. Moreover, Tang et al. observed that mTOR hyperactivation in mice led to ASD-like behaviors and spine pruning deficits in cortical projection neurons, indicating that autophagy is essential for synaptic pruning, and that defects in autophagy contribute to ASD phenotypes (Tang et al. 2014).

Overall, these studies show that autophagic machinery is vital for both synapse formation and pruning during development. However, many questions remain. For example, does the autophagic machinery regulate SV clustering via protein degradation or another mechanism? Are other aspects of synapse formation also regulated by this machinery? Most importantly, how are autophagy components temporally and spatially regulated to facilitate SV clustering in certain contexts, and synapse pruning or SV degradation in others? Answers to these questions will lead to a deeper understanding of how autophagy functions throughout neurodevelopment to facilitate synapse formation and elimination.

Autophagic Cargo at the Synapse

Mitochondria

Mitochondria can become depolarized and dysfunctional following damage to mitochondrial DNA by reactive oxygen species (ROS) (Guo et al. 2013). The term ‘mitophagy’ was coined by Rodriguez-Enriquez and colleagues in 2006 to describe the selective autophagy of damaged or excessive mitochondria (Rodriguez-Enriquez et al. 2006). Since then, mitophagy has been shown to be essential for the removal of damaged mitochondria in all eukaryotic cell types, including neurons (Ding and Yin 2012; Tang et al. 2014). Mitochondria are critical sources of

ATP at synapses, and several studies indicate that deficient mitophagy leads to synaptic dysfunction (Ebrahimi-Fakhari et al. 2016; Haddad et al. 2013). Recent studies implicate the

Parkinson’s disease (PD)-related proteins PTEN-induced putative kinase (PINK1) and Parkin, an

E3 ubiquitin ligase, as critical mediators of mitophagy in axons and presynaptic terminals of hippocampal neurons (Ashrafi et al. 2014). These proteins were previously shown to work in concert to target mitochondria for degradation, as PINK1 localizes to the mitochondrial

membrane and upon phosphorylation recruits Parkin to damaged mitochondria (N. Matsuda et al.

2010; Narendra et al. 2010; Vives-Bauza et al. 2010). Parkin in turn ubiquitinates mitochondrial membrane proteins to promote recruitment of the autophagic machinery (Mouton-Liger et al.

2017; X. Wang et al. 2011). However, more recent studies call into question whether the

PINK1/Parkin pathway is essential for triggering the removal of damaged mitochondria from axons and presynaptic boutons. For example, Sung and colleagues found that Parkin deficiency in Drosophila led to the accumulation of abnormal mitochondria in cell bodies, rather than in axons or presynaptic boutons as would be expected if PINK1/Parkin are required for initiating the clearance and retrograde transport of these organelles (Sung et al. 2016). Moreover, another study showed that the removal of stressed mitochondria from axons of hippocampal neurons required the retrograde motor protein syntaphilin but not Parkin (M. Y. Lin et al. 2017), suggesting a Parkin-independent mechanism for detecting mitochondrial dysfunction and mediating its retrograde transport prior to mitophagy. Additional evidence shows that basal mitophagy is unaffected in Pink1 or parkin null Drosophila, although these flies exhibit locomotor defects and dopaminergic neuron loss, demonstrating that PINK1 and Parkin are not essential for in vivo mitophagy, or that their function is cell-type specific (J. J. Lee et al. 2018).

Synaptic Vesicles

Autophagy has also been found to mediate the degradation of synaptic vesicles (SVs) in some circumstances. Notably, Hernandez and colleagues found that in dopaminergic neurons, stimulating autophagy with rapamycin decreased SV number and evoked-dopamine release, while blocking autophagy via ATG-7 knockout increased dopamine release (Hernandez et al.

2012). These findings suggest that, at least in dopaminergic neurons, induction of autophagy can

stimulate SV degradation, while loss of this pathway leads to increased SV release and/or number. However, other studies in Drosophila and mammalian hippocampal neurons suggest that the endolysosomal degradative pathway mediates basal and activity-dependent turnover of

SV proteins (Sheehan et al. 2016; Uytterhoeven et al. 2011, 2017). A synthesis of these studies suggests that although activation of autophagy can stimulate SV degradation and regulate neurotransmitter release under some conditions, autophagy may not be the primary mechanism of SV protein degradation. It is also worth noting that the autophagy and endolysosomal pathways are linked, as autophagosomes fuse with late endosomes prior to lysosomal degradation (Berg et al. 1998; Liou et al. 1997; Fader and Colombo 2008). Thus, it can be difficult to parse the roles of the autophagy and endolysosomal pathways in the degradation of specific cargoes, and in the case of synaptic vesicle degradation it is possible that the two pathways work synergistically.

Regardless of the specific substrates, considerable evidence suggests that presynaptic autophagy is tightly regulated and, distinct from basal autophagy, activated only under certain conditions (Vijayan and Verstreken 2017). Indeed, several recent studies have begun to identify the molecular machinery that activates and inhibits autophagy in presynaptic terminals. Okerlund and colleagues report that the active zone cytomatrix proteins Bassoon and Piccolo are negative regulators of presynaptic autophagy in hippocampal neurons (Okerlund et al. 2018). Following up on previous findings that Bassoon/Piccolo double knockdown (DKD) triggered the progressive loss of SV pools by stimulating degradative pathways (Waites et al. 2013), the authors observed an accumulation of LC3-positive and autophagosome-like vesicles in presynaptic DKD boutons (Okerlund et al. 2018). They subsequently demonstrated that Piccolo and Bassoon regulate autophagy via their suppression of ATG-5, an E3 ubiquitin ligase-like

enzyme essential for coupling LC3 to the membrane of nascent autophagosomes. Both active zone proteins directly interact with ATG-5, and its knockdown rescues the increase in LC3- positive structures and concurrent decrease in SV pools seen in DKD neurons (Okerlund et al.

2018). These data indicate that Bassoon/Piccolo act to restrain presynaptic autophagy, and that dysregulated autophagy due to the loss of these proteins stimulates the degradation of SVs.

However, it is not yet clear whether autophagy is specifically regulated by Bassoon/Piccolo or induced by a broader neuronal stress response following the loss of one or both of these critical synaptic structural determinants. More work is needed to clarify the roles of these proteins in regulating presynaptic autophagy and other degradative pathways (e.g. endolysosomal, ubiquitin- proteasome), as well as the involvement of autophagy in presynaptic vesicle degradation under basal conditions.

Additional insights about the molecular regulation of presynaptic autophagy come from two recent studies in the Verstreken lab (Soukup et al. 2016; Vanhauwaert et al. 2017). Soukup and colleagues showed that autophagic factors are recruited to presynaptic boutons by the endocytic protein EndophilinA (EndoA) following its phosphorylation by LRRK2, a Parkinson’s disease-related kinase (Soukup et al. 2016). Vanhauwaert et al. found that synaptojanin, a presynaptic lipid phosphatase also important for SV endocytosis, is similarly crucial for autophagy. In Drosophila synapses and neurites of iPSC-derived neurons with mutant synaptojanin, nascent autophagosomes accumulated Atg18a on their membranes and could not mature into functional autophagosomes (Vanhauwaert et al. 2017). Moreover, Parkinson’s disease-associated synaptojanin mutations caused dopaminergic neuron loss in Drosophila

Vanhauwaert et al. 2017), linking the dysregulation of presynaptic autophagy to

neurodegenerative disease, and highlighting the importance of this process for nervous system health.

Postsynaptic AMPA receptors

Across the synaptic cleft, autophagy also mediates the degradation of postsynaptic cargo, including neurotransmitter receptors. Some of the first evidence for this came from Rowland and colleagues, who showed that GABAA receptors colocalize with autophagosome markers at the C. elegans motor neuron/body wall junction in the absence of neuronal activity, and that postsynaptic GABAA activity is attenuated in an autophagy-dependent manner (Rowland et al. 2006).

Interestingly, the authors also found that this autophagic degradation was selective, as GABAA but not acetylcholine receptors localized to autophagosomes (Rowland et al. 2006). Matsuda et al. subsequently observed that hippocampal and cerebellar AMPA receptors were mistargeted to axonal autophagosomes in the absence of the trafficking adaptor complex AP-4, instead of being delivered to the dendritic plasma membrane (S. Matsuda et al. 2008). Although this study did not demonstrate autophagy of AMPA receptors in dendrites, a later study reported more direct evidence of AMPA receptor degradation at synapses through the autophagy-lysosomal pathway.

Specifically, Shehata and colleagues found that high potassium stimulation or chemical stimuli to induce long-term depression (LTD) briefly increased autophagosome numbers in dendritic shafts and spines, an effect that was partially NMDA receptor-dependent (Shehata et al. 2012).

Moreover, the levels of AMPA receptor subunit GluA1 decreased following chemical LTD, and this degradation was partially mitigated by autophagy inhibitors. These results indicate that postsynaptic autophagy is regulated by neuronal activity, and at least partially responsible for the degradation of AMPA receptors following LTD (Shehata et al. 2012), suggesting a possible role

for autophagy in synaptic plasticity mechanisms. However, given the preponderance of evidence demonstrating that AMPA receptors are degraded through the endolysosomal pathway (Ehlers

2000; Zheng et al. 2015; Schwarz, Hall, and Patrick 2010) and ubiquitin proteasome system

(UPS) (Ferreira et al. 2015; Goo, Scudder, and Patrick 2015), it will be important to clarify whether autophagy works in parallel with these other pathways, or is stimulated only during specific circumstances, e.g. induction of LTD. Despite many lingering questions, it is clear that autophagy mediates the degradation of important synaptic substrates (Figure 1), and that future studies are needed to disentangle and clarify the many roles it plays on both sides of the synaptic cleft.

Figure 1. Proposed regulators and cargo of synaptic autophagy. In the presynaptic terminal, autophagosome formation is promoted by EndophilinA (EndoA) following its phosphorylation by

LRRK2. Conversely, the presynaptic structural proteins Bassoon and Piccolo inhibit autophagic degradation of synaptic vesicles via negative regulation of ATG-5. Mitophagy is induced when PINK1 is phosphorylated and recruits Parkin to damaged mitochondria. Postsynaptically, AMPA receptors are degraded via autophagy following induction of LTD.

Outstanding Questions

Given the distinct molecular compositions and features of the juxtaposed pre- and postsynaptic compartments, one important unresolved question is whether pre and post-synaptic autophagosomes have unique characteristics and/or regulatory mechanisms. A recent study by

Maday and colleagues supports this idea, as the authors find that autophagosomes originating in axons of hippocampal neurons exhibit distinct characteristics from those originating in the somatodendritic region (Maday and Holzbaur 2016). The former appear overwhelmingly

‘mature,’ i.e. positive for the late endosome/lysosome marker LAMP1, once they enter the somatodendritic compartment following their retrograde axonal transport, while the latter are less mature and less mobile overall (Maday and Holzbaur 2016). These findings suggest that neuronal and/or synaptic compartmentalization creates molecularly distinct pools of autophagosomes that exhibit different dynamic behavior and functional properties. Additional work will be needed to determine whether these autophagosomes are indeed molecularly distinct and/or responsive to compartment-specific regulatory stimuli.

This possibility of distinct spatially and conditionally regulated autophagosomes dovetails with mounting evidence that there are multiple classes of LC3-positive structures.

Although LC3 is considered a specific and reliable marker of autophagosomes, it is worth noting that LC3 was first identified in rat neurons as the light chain of microtubule-associated proteins

MAP1a and MAP1b, components of the neuronal cytoskeleton (Mann and Hammarback 1994).

Its role in autophagy was recognized only later by Kabeya and colleagues (Kabeya 2000). In keeping with the potentially versatile role of LC3, Kononenko et al. reported that a subset of

LC3-positive structures undergoing retrograde transport from presynaptic terminals exhibited colocalization with the endocytic adaptor AP-2 and with BDNF-activated TrkB receptors, which are traditionally found in signaling endosomes (Kononenko et al. 2017). One interpretation of these data is that presynaptic autophagosomes can merge with signaling endosomes, or that these retrograde cargoes are co-transported. Another possibility is that LC3 associates with multiple vesicle types, including autophagosomes, signaling endosomes, and potentially other structures.

Figure 2. Diverse membrane structures observed at Bassoon/Piccolo DKD synapses. A) Electron micrograph of a presynaptic terminal expressing VAMP2-HRP to label synaptic vesicles (red arrow). ‘M’ denotes mitochondria. B) Examples of structures found at DKD synapses, shown to have increased autophagy and LC3-positive vesicles. These structures include vacuoles (arrowheads), putative double- membrane autophagosome (red arrow), multivesicular bodies (green arrows), as well as structures that are difficult to characterize due to electron-dense VAMP2-HRP labeling (yellow arrows).

Interestingly, it was observed by Waites, Okerlund, and colleagues that Bassoon/Piccolo DKD boutons, shown to be LC3-immunopositive, contained multiple heterogeneous membrane structures, most of which did not resemble classical autophagosomes (Waites et al. 2013;

Okerlund et al. 2018) (Figure 2). It is also worth noting that the other mammalian Atg8 homologs, namely GABARAP and GATE-16, likewise exhibit many functions at the synapse, some unrelated to autophagosome formation. GABARAP in particular has been shown to be involved in the trafficking of GABA-A receptors to the neuronal plasma membrane (Leil et al.

2004), and has numerous binding partners that are involved in a diversity of dynamic processes underlying synapse formation, maintenance, and plasticity (Mohrlüder, Schwarten, and Willbold

2009). Together, these studies indicate how much there still is to learn about protein trafficking and degradative pathways at the synapse.

ESCRT Pathway

ESCRT Pathway Components and Mechanics

The endosomal sorting complex required for transport (ESCRT) pathway is required for the endolysosomal degradation of membrane proteins in eukaryotic cells. ESCRT machinery comprises a series of protein complexes termed ESCRT-0, I, II, III, and VPS4/Vta1, which together form an ancient system for membrane remodeling (Figure 3) (Hurley 2015). Each complex is conserved at least in part from yeast to mammals, and plays a role in the formation of multivesicular bodies, pre-degradative endosomal structures characterized by cargo-carrying intraluminal vesicles. The loss or dysfunction of almost every ESCRT component results in

endosomal dysfunction, usually the formation of class-E structures in yeast, or enlarged endosomes in mammalian cells (Schuh and Audhya 2014).

Figure 3. The ESCRT Pathway. ESCRT Complexes ESCRT-0,I,II,III and Vps4/Vta1 (not shown) are recruited sequentially to the endosomal membrane to sequester ubiquitinated membrane protein cargo and insert it into intralumenal vesicles, ultimately creating a multivesicular body (MVB).

ESCRT-0 is the first component of the pathway and is necessary to cluster ubiquitinated cargo in the endosomal membrane as well as recruit later ESCRT complexes. Comprising two proteins, Hrs (Hepatocyte growth factor-regulated tyrosine kinase substrate; 777 amino acids) and STAM (Signal transducing adaptor molecule; 540 amino acids; isoforms STAM1 or

STAM2) (Figure 4), ESCRT-0 is thought to exist as a 2:2 heterotetramer on the endosomal membrane (Mayers et al. 2011; Schuh and Audhya 2014). Hrs localizes to the endosome through interactions between its Fab1, YOTB/ZK632.12, Vac1, and EEA1 (FYVE) domain and

phosphoinositol-3-phosphate, and between its coiled-coil (CC) domain and the membrane

(Raiborg et al. 2001). It binds ubiquitin with a 127 M affinity in vitro using a double-sided ubiquitin interacting motif (DUIM), and clusters ubiquitinated cargo on the endosomal membrane, localizing it to clathrin-rich domains via its C-terminal clathrin-binding (CB) domain

(Schuh and Audhya 2014). Hrs and STAM interact at their CC domains. STAM also binds ubiquitin, albeit more weakly, with UIM and VHS domains (Bache, Raiborg, et al. 2003).

Additionally, STAM has an SH3 domain that has been shown to interact with deubiquitinases

(DUBs) (Schuh and Audhya 2014). ESCRT-0 thus recognizes and clusters ubiquitinated cargo into specific membrane microdomains that become the sites of intraluminal vesicle formation. It also recruits the next component in the pathway, the ESCRT-1 complex, via an interaction between the Hrs PxxP domain and the ubiquitin E2 variant (UEV) domain of ESCRT-1 protein

TSG101 (Bache, Brech, et al. 2003; Schuh and Audhya 2014).

Figure 4. Relevant domains of ESCRT-0 proteins Hrs and STAM. Abbreviations are as follows: Clathrin Binding (CB), Coiled-Coil (CC), Double-Sided Ubiquitin-Interacting Motif (DUIM), Fab-1, YGL023, Vps27, and EEA1 (FYVE), P= Proline, x= any amino acid (PxxP), Src homology-3 (SH3), Ubiquitin Interacting Motif (UIM), Vps27p, Hrs, and STAM (VHS).

ESCRT-I comprises four conserved proteins: TSG101 (390aa), Mvb12B, Vps37, and

Vps28. Unlike ESCRT-0, ESCRT-I does not stably associate with the membrane and relies on its interaction with ESCRT-0 to target the endosomal membrane. The UEV domain of TSG101

weakly binds ubiquitin with ~510 M affinity (Bilodeau et al. 2003; Katzmann, Babst, and Emr

2001). This interaction does not appear to be necessary for cargo sorting, but along with the ubiquitin-binding motif present in ESCRT-II, may represent a sort of outer failsafe for retaining ubquitinated cargo in the area (Schuh and Audhya 2014).

ESCRT-II similarly does not stably associate with endosomal membranes, but rather forms a heterotetramer in solution, consisting of its three protein components: Vps22, Vps36 and two copies of Vps25 (Im and Hurley 2008). Vps36 contains a Gram-like ubiquitin-binding in

Eap45 (GLUE) domain that binds ubiquitin and exhibits a weak preference for P13P. ESCRT-I and ESCRT-II may be somewhat involved in the creation of membrane curvature, but the main role of ESCRT-II seems to be to recruit ESCRT-III to the membrane and seed its polymerization via interactions between Vps25 and ESCRT-III component Vps20 (Schuh and Audhya 2014).

ESCRT-III assembles only on lipid bilayers, and does so via the sequential recruitment of its four components: Vps20, Vps32, Vps24 and Vps2 (Babst et al. 2002). Vps20 appears to be autoinhibited in the cytosol, but its recruitment by Vps25 to the endosomal membrane allows it to form an open conformation that promotes the assembly of ESCRT-III polymers. All components of ESCRT-III exhibit a similar structure, and are thought to adopt a 4-helix bundle configuration, with a fifth helix connected by a flexible linker (Muzioł et al. 2006; Bajorek et al.

2009). These helical bundles polymerize into spiraling chains that deform the membrane, causing it to bud inward and form intraluminal vesicles carrying the cargo clustered by the earlier

ESCRT components (Schuh and Audhya 2014).

During the budding of nascent vesicles into the interior of the MVB, the final component of the ESCRT pathway, the Vps4 complex, gets to work. This complex consists of Vps4 and its adaptor protein, Vta1. Vps4 is a type 1 AAA ATPase that forms dodecameric or tetradecameric

complexes with a central pore (Babst et al. 1998) that directly contacts ESCRT-III subunits and triggers their disassembly from the membrane, releasing ESCRT-III. In this model, the stepwise removal of ESCRT-III subunits from the polymer results in an ever-shrinking vesicle neck, which eventually closes to pinch off the vesicle inside the MVB (Babst et al. 1998; Schuh and

Audhya 2014).

In order to create an MVB, this process is repeated until all of the ubiquitinated cargoes on the endosomal membrane are trapped in intraluminal vesicles (Wenzel et al. 2018). Although a single ubiquitin attached to a membrane protein is sufficient to activate this pathway, the affinity of ESCRT machinery for polyubiquitin chains is significantly higher (Ren and Hurley

2010). In cells, the interactions of ESCRTs with ubiquitinated membrane proteins happens in equilibrium with those of DUBs, which remove ubiquitin from potential cargo, in order to establish the correct balance of membrane protein trafficking for degradation versus recycling.

Mutation or deletion of ESCRT components, or an increase in DUB activity, will slow or halt the degradation of endolysosomal cargo (Shenoy and Lefkowitz 2003). In contrast, an increase in E3 ligase activity or a decrease in DUB activity will speed degradation (McCullough, Clague, and

Urbé 2004). However, some level of DUB activity is necessary for the creation of MVBs, as cargo must be deubiquitinated in order to release earlier ESCRTs and enable ILV formation

(Henne, Buchkovich, and Emr 2011). This delicate balance must be maintained by all cells, and its disruption can have significant implications for cell health.

Regulation of Membrane Protein Degradation

The endolysosomal pathway mediates the degradation of ubiquitinated membrane- associated proteins by orchestrating their recognition, capture, and sorting into multivesicular

bodies for delivery to lysosomes. All of these activities are carried out by the endosomal sorting complex required for transport (ESCRT) machinery described above. The ESCRT pathway is necessary for the homeostatic degradation of myriad membrane proteins, and this process is regulated by a complex variety of factors that differ depending on cell type (Schuh and Audhya

2014). One of the best characterized cargoes of the ESCRT pathway is the epidermal growth factor receptor (EGFR). An intricate network of E3 ligases and DUBs regulates the ubiquitination of EGFR, most often with K63 polyubiquitination. Ubiquitin chains can be linked via different lysines, and it has been shown that different linkages result in divergent regulation in mammalian cells. Of the eight ubiquitin linkages, K63 linkages are one of the most common, and have been shown to play a proteasome-independent role in the cell. K63 linkages are instead involved in processes including endocytosis and selective autophagy (Ohtake et al. 2018;

Husnjak and Dikic 2012), making them a prime candidate for the regulation of ESCRT cargoes like EGFR. Depletion of AMSH (associated molecule with SH3 domain of signal transducing adaptor molecule), a DUB that interacts directly with ESCRT components, speeds the degradation of EGFR, likely by slowing its rate of deubiquitination (McCullough, Clague, and

Urbé 2004). However, another DUB involved in this process, UBPY (ubiquitin isopeptidase Y), has been shown to speed or slow EGFR degradation in different contexts (Bowers et al. 2006;

Mizuno et al. 2005; Row et al. 2006), highlighting the complexity of the role of DUBs in ESCRT regulation. In terms of ubiquitination, both the Cbl and Nedd families of E3 ligases appear to play a role in stimulating the degradation of ESCRT cargo (Schuh and Audhya 2014). The many players involved in regulating ESCRT-mediated protein degradation, and the delicate balance of their actions even in simple eukaryotic cell types, underscores how much there is still to learn about ESCRT regulation in neurons.

ESCRT Pathway in Neurons

Degradation of Synaptic Vesicle Proteins

The inability to clear damaged or misfolded SV proteins can precipitate synaptic dysfunction and neurodegeneration (Bezprozvanny and Hiesinger 2013; Esposito, Ana Clara, and Verstreken 2012; Hall et al. 2017). However, the elaborate morphology of neurons creates unique spatial challenges for SV protein clearance and degradation. First, lysosomes - the degradative organelles for SV membrane proteins - are primarily localized to neuronal cell bodies, requiring retrograde transport of SV proteins destined for degradation. Second, the machinery responsible for packaging and delivering SV proteins to lysosomes must be anterogradely transported to presynaptic boutons. In addition to these spatial challenges, neurons face temporal challenges in transporting degradative machinery in response to stimuli such as synaptic activity. For instance, degradative organelles (proteasomes, lysosomes, and autophagosomes) undergo activity-dependent recruitment into the postsynaptic compartment as part of the mechanism for synaptic plasticity (Bingol and Schuman 2006; Goo et al. 2017;

Shehata et al. 2012), and must be rapidly mobilized to these sites. Similarly, the turnover of SV and other presynaptic proteins is stimulated by neuronal activity (Sheehan et al. 2016;

Truckenbrodt et al. 2018), requiring the local presence of degradative machinery to facilitate clearance of old and damaged proteins. However, very little is known about whether or how neurons regulate the transport and delivery of this machinery to presynaptic terminals.

Previous work from our group and others has demonstrated that the degradation of SV membrane proteins requires the endosomal sorting complex required for transport (ESCRT) pathway (Sheehan et al. 2016; Uytterhoeven et al. 2011). Sheehan et . al. showed that a

knockdown of ESCRT-0 protein Hrs slows the degradation of synaptic vesicle proteins, including SV2 and VAMP2. These findings, and the importance of the ESCRT pathway in neurons, is underscored by the clear connection between ESCRT dysfunction and neurodegeneration.

Role in Neurodegeneration

Studies in both animal models and humans have made it clear that the ESCRT pathway is vital for maintaining neuronal health. Mutation or dysfunction of various ESCRT-associated proteins leads to neurodegeneration in animal models (Tamai et al. 2008; J. A. Lee et al. 2007;

D. Lee et al. 2019). ESCRT dysfunction is also linked to frontotemporal dementia (FTD)

(Skibinski et al. 2005; J. A. Lee et al. 2007), amyotrophic lateral sclerosis (ALS) (Parkinson et al. 2006) and infantile-lethal epilepsy (Hall et al. 2017) in humans. Moreover, the ESCRT pathway is essential for endosomal protein trafficking, and is implicated in the etiology of

Alzheimer’s and Parkinson’s diseases (Vaz-Silva et al. 2018; Borland and Vilhardt 2017;

Vagnozzi and Praticò 2019; Feng et al. 2020; Spencer et al. 2014, 2016). Yet despite the critical role of the ESCRT pathway in protein degradation and other cellular processes, as well as its links to neurodegenerative disease, little is known about its substrates, dynamic behavior, or mechanisms of localization and regulation in neurons.

A previous study from our lab demonstrated that a subset of SV proteins are degraded in an activity-dependent manner, and that this process is initiated by the activation of small GTPase

Rab35 and the recruitment of its effector, ESCRT-0 component Hrs (Sheehan et al. 2016).

Interestingly, neuronal activity significantly increases the localization of Hrs to axons and SV pools, suggesting that ESCRT-0 proteins could be transported to SV pools in an activity-

dependent manner to catalyze SV protein degradation. However, it is not clear how ESCRT-0 proteins are transported into and out of synapses, leaving the questions of spatial and temporal regulation of neuronal ESCRTs unanswered.

Endosomal Transport Dynamics

Mechanics of Axonal Transport

Axonal transport of cargoes anterogradely towards presynaptic boutons and retrogradely towards the soma is crucial for maintenance of synaptic health. The basics of axonal transport are the same for all cargoes. Anterograde transport is mediated by kinesins, and in neurons motors from the kinesin-1, kinesin-2, and kinesin-3 families all contribute to anterograde axonal transport. Retrograde transport is mediated by the singular dynein motor and facilitated by a wide variety of dynein adaptors. All motor proteins hydrolyze ATP to move progressively down a microtubule, towards the plus-end for kinesins, and the minus-end for dynein (Maday et al.

2014). However, a more granular look at the dynamics and players involved in transport of individual cargoes shows that the regulation of axonal transport is complex and incredibly specific to individual cargoes, be they mitochondria, cytosolic proteins, synaptic vesicle precursors, mRNA, signaling endosomes, late endosomes, BDNF vesicles, early endosomes, or autophagosomes (Maday and Holzbaur 2016; Maday et al. 2014; Maday, Wallace, and Holzbaur

2012; Obashi and Okabe 2013; M.-Y. Lin and Sheng 2015; K. Zhang et al. 2013; Bentley et al.

2015; Goto-Silva et al. 2019). It is therefore clear that to understand the axonal regulation of an organelle or cargo, it is first necessary to examine its specific transport dynamics and the motors involved.

Autophagosome Transport

It is evident from the literature that autophagosomes initiate in the distal axon, and are then transported retrogradely down the axon by dynein, acidifying into autolysosomes as they get to the soma (Maday, Wallace, and Holzbaur 2012; Maday and Holzbaur 2016; Kulkarni and

Maday 2018a; Maday et al. 2014; Shehata et al. 2012; T. Wang et al. 2015; Borland and Vilhardt

2017). Although autophagosomes interact with kinesin-1 and show some bidirectional movement immediately after formation in the distal axon, >80% eventually travel retrogradely to the soma

(Maday et al. 2014). The major protein marker used to identify autophagosomes in imaging assays is LC3 (Maday and Holzbaur 2016; Maday, Wallace, and Holzbaur 2012), the lipidated form of which (LC3-II) is a widely cited, albeit slightly controversial (Kononenko et al. 2017;

Okerlund et al. 2018; Waites et al. 2013), marker of autophagosomes (Kabeya 2000; Tanida,

Ueno, and Kominami 2004).

Endosomal Transport

Endosomes exhibit a much more complex axonal transport pattern. Late endosomes are generally marked by Rab7 and LAMP1 in the axon, and show limited catalytic activity, distinguishing them from the mature lysosomes found in the soma (K. Zhang et al. 2013;

Deinhardt et al. 2006; Cheng et al. 2018; Kulkarni and Maday 2018b; Yap et al. 2018). Axonal late endosomes copurify with kinesin-1, 2, and dynein (Maday et al. 2014), and exhibit bidirectional transport. About half of these structures exhibit bidirectional motility characterized by frequent direction switching, and the other half display more progressive motility in either the anterograde or retrograde direction (Maday et al. 2014; Hendricks et al. 2010), although studies

have highlighted the particular importance of their retrograde transport (Ferguson 2018).

Interestingly, it has also been shown that autophagosomes may fuse with late endosomes in order to be transported in the retrograde direction (Cheng et al. 2015). Early endosomes, marked in the axon by Rab5, also exhibit a complex bidirectional motility, with biases towards anterograde or retrograde motility emerging in the presence of additional markers (Olenick, Dominguez, and

Holzbaur 2018; Huckaba et al. 2011; Goto-Silva et al. 2019). These studies make it clear that in axons, Rab5 marks a diverse pool of vesicles ranging from signaling endosomes to classical early endosomes, all with unique patterns of motility. Thus, our understanding of the intricacies of endosomal axonal dynamics remains far from complete.

Chapter 2: Materials and Methods

Cell Culture

Neurons Hippocampal neurons were prepared from E18 Sprague Dawley rat embryos of both sexes using a modified Banker culture protocol (Kaech & Banker, 2006, Nature Protocols;

Sheehan et al., 2016, The Journal of Neuroscience; Waites et al., 2009, The Journal of

Neuroscience). Neurons were dissociated in TrypLE Express (ThermoFisher/Invitrogen) for

20min, washed with Hanks Balanced Salt Solution (Sigma), and plated in Neurobasal medium

(ThermoFisher/Invitrogen) with N21 supplement (R&D Systems) and Glutamax

(ThermoFisher/Invitrogen) at a density of 250,000 neurons per well (12 well plates), coverslip

(22 x 22 mm square) or 35mm glass-bottom dish (MatTek), or at a density of 100,000 neurons per microfluidic chamber. Coverslips and dishes were pre-coated with 0.25 mg/ml poly-L-lysine

(PLL) (Sigma-Aldrich) in 100 mM boric acid (pH 8.0) for 24 h before plating.

Cell Lines Neuro2a (N2a) neuroblastoma cells (ATCC CCL‐131) and HEK293T cells (Sigma) were cultured in DMEM‐GlutaMAX (ThermoFisher/Invitrogen) with 10% FBS (Atlanta Biological) and Antibiotic‐Antimycotic (ThermoFisher/Invitrogen) and kept at 37°C in 5% CO2.

Microfluidic Chambers

Microfluidic molds and chambers were prepared as described in (Birdsall et al., 2019).

Master molds were prepared in a nanofabrication facility by repeatedly flooding SU-8 2050 over a silicon wafer and subjecting it to heat and UV light through a mask that outlines the mold

pattern (this step was completed by Lisa Randolph, Jose Martinez, and Ethan McCurdy in the lab of Ulrich Hengst at Columbia University). Once the molds were ready, they could be used repeatedly to construct chambers out of poly(dimethylsiloxane) (PDMS).

Chambers were prepared by pouring ~30 ml of PDMS mixture (QSil 216) into molds, then baking. QSil 216 was weighed out in a 10:1 ratio of silicone polymer: curing agent in disposable plastic cup and mixed thoroughly for 2-5 min. The PDMS polymer solution was then vacuum mixed in a vacuum desiccator for 1 hour or until bubbles are mostly gone. The master mold was cleaned from dust and particles using tape, then the solution was carefully poured into the mold, and allowed to sit at room temperature for 2 min. The mold was then cured in a level laboratory oven for at least 3 h to overnight at 70°C. A razor blade was then used to cut around the inner edge of the mold and slowly remove the cured polymer piece. Individual chambers were then cut out using a razor blade, and channels were punched out channels using steel punch.

To prepare chambers in dishes for cell seeding, glass-bottom dishes were incubated in

PLL solution for >12 hours at 37°C, then washed 3x with deionized sterile water (dH2O) and dried. Chambers were sterilized by immersing briefly in 70% ethanol and dried fully in a laminar hood before use. The sterile, dried chambers and dishes were assembled by placing chamber lightly on glass, grooved side down. After assembly, channels were filled with warmed neuron culture media. Chambers were assembled chambers 3 - 24 h before seeding and placed in a humidified incubator at 37°C with 5% CO2. Next, a E18 rat hippocampal dissociated cell suspension was prepared (see Cell Culture), with an approximate density of 10,000 cells/ml of

Plating media. Media was aspirated from chamber wells, cells were seeded into the two top left wells of a tripartite chamber, 70,000 cells in the upper left well, and 30,000 in the upper middle well (Figure 5). Chambers were placed back in incubator for 30 min to 1 h, then the remaining

wells were filled with Plating media. Plating media was removed and replaced with Neurobasal containing antibiotics 24 hours later.

Figure 5. Schematic diagram of microfluidic chambers and lentiviral transduction.

DNA Constructs

pVenus-2xFYVE was a gift from F. Polleux and was subcloned into pFUGWm vector

(Waites et al. 2013) at MfeI/NheI sites; pFU-RFP-Hrs-W, pFU-STAM-mCh-W, pFU-

EGFP/mCh-Rab35, and pFU-EGFP/mCh-Rab5 were described in (Sheehan et al. 2016)); pFU-

EGFP-Synapsin-W was described in (Leal-Ortiz et al. 2008); and pBa-eGFP-flag-BicD2594-

FKBP (Addgene plasmid #63569; http://n2t.net/addgene:63569; RRID:Addgene_63569), pBa-

FRB-3myc-KIF13A tail 361-1749 (Addgene plasmid #64288; http://n2t.net/addgene:64288;

RRID:Addgene_64288), and pBa-FRB-3myc-KIF13B tail 442-1826 (Addgene plasmid #64289; http://n2t.net/addgene:64289; RRID:Addgene_64289) were gifts from Gary Banker and Marvin

Bentley (Jenkins et al. 2012; Bentley et al. 2015; Bentley and Banker 2015). Full-length human

Hrs (NCBI accession #D84064.1) was synthesized at Genewiz and cloned into pFUGWm vector at BsrGI/XhoI sites to create pFU-EGFP-Hrs-W, and EGFP replaced with Halotag at the

AgeI/BsrGI sites to create pFU-Halo-Hrs-W. To create pFU-EGFP-R183A Hrs-W, an 868- nucleotide fragment of Hrs containing the R183A mutation was synthesized at Genewiz and subcloned into the BsrGI/PshAI sites. shRNA scrambled control and KIF13A knockdown constructs (shCtrl, sh13A) were purchased from Origene (catalog #TL706020). The 29mer shRNA (lower-case letters) and surrounding loop sequences (capital letters)

(shCtrl:CGGATCGgcactaccagagctaactcagatagtactTCAAGAG agtactatctgagttagctctggtagtgcTTTTT; sh13A:CGGATCGagccagacctctatgatagcaaccatcag

TCAAGAGctgatggttgctatcatagaggtctggctTTTTTT) were then re-synthesized and subcloned into pFUGW U6 vector (Waites et al. 2013) at EcoRI/PacI sites.

Pharmacological Treatments

Pharmacological agents were used in the following concentrations and time courses: cycloheximide (Calbiochem, 0.2 µg/µl, 24 h), bicuculline (Sigma, 40 µM, 5 min-2 h), 4- aminopyridine (Tocris Bioscience, 50 µM, 5 min-2 h), D-AP5 (D-(-)-2-Amino-5- phosphonopentanoic acid; Tocris Bioscience, 50 µM, 1 h), CNQX (6-Cyano-7-nitroquinoxaline-

2,3-dione; Tocris Bioscience, 10 µM, 1 h), TTX (Tetrodotoxin; Tocris Bioscience, 0.5 µM, 2 h),

PR-619 (LifeSensors, 50 µM, 2 h), Rapamycin analogue linker drug (AP21967, Clontech, 100 nM, 3 h), SAR405 (Millipore Sigma, 1 µM, 2 h). Unless otherwise indicated, all other chemicals were purchased from Sigma.

Evaluation of Neuronal Health

Cell health in microfluidic chambers was evaluated using NucRed Dead 647 reagent to identify cells with compromised membranes along with membrane-permeant NucBlue™ Live

ReadyProbes™ Reagent (both ThermoFisher) to evaluate the total number of cells. Neurons were treated with these reagents for 15 min then immediately imaged as described (see Live

Neuron Imaging).

Lentiviral Transduction and Neuronal Transfection

Lentivirus was produced using HEK293T cells as previously described (Birdsall et al.,

2019; Sheehan et al., 2016). Briefly, HEK293T cells were plated in HEK cell media in a culture- treated 10 cm dish, such that they would be 60-70% confluent at transfection. Media was changed prior to transfection. For each plate, a transfection mixture was set up in a 1.5 ml tube, containing: 1 ml DMEM (serum-free), 10 μg FUGW plasmid (containing the desired ESCRT construct), 7.5 μg delta P (packaging plasmid), 5 μg VsVg (envelope gene). Tubes were vortexed then spun down briefly before the addition of 35-40 μL of CalFectin reagent/tube. Tubes were incubated for 15 min at room temperature, then the mixture was added dropwise to 10 cm plate of HEK cells. 12-18 h after transfection, HEK cell media was removed and replaced with 8 mL of Neurobasal media. Virus was collected and filtered with a 0.45 μm filter 24 h after media change. Neurons were transduced with 50-150 µL of lentivirus on 7 DIV (days in-vitro), and imaged or fixed on 13-15 DIV. Lentivirus was fluidically isolated in chambers as previously described (Birdsall et al., 2019). Briefly, ~50% of the media from the wells to be infected was removed and placed in adjacent wells, then virus was infected into the reduced-volume wells to contain it within those compartments (Figure 5).

Neurons were transfected as described in (Sheehan et al. 2016) using Lipofectamine 2000

(ThermoFisher/Invitrogen) on 9 DIV. Briefly, for each 22x22 mm coverslip or 35 mm dish, 2.5

µL Lipofectamine was incubated with 62.5 µL Neurobasal for 5 min, combined with 2.5 µg

DNA diluted into 62.5 µL Neurobasal for 15 min, then added to neurons for 1 h at 37°C.

Neurons were transfected in Neurobasal medium containing 50 µM AP5 and 10 µM CNQX, washed 3X in this medium, and then returned to their original medium following transfection.

Electron Micrograph Analysis of Endosomal Structures

Electron microscopy of hippocampal synapses was performed by Glebov and Burrone as described (Glebov et al. 2016). Presynapses were defined visually by the presence of synaptic vesicle pools. Endosomal structures were defined as single-membraned structures with a diameter >80 nm. Multilamellar structures were defined as structures with more than one visible membrane. Synaptic area and endosomal/multilamellar structure size were measured using FIJI software.

Correlative Light and Electron Microscopy

For correlative fluorescence and electron microscopy experiments (performed by Yuuta

Imoto in the lab of Shigeki Watanabe at John’s Hopkins University), mouse hippocampal neurons were cultured on sapphire disks (Technotrade #616-100) that were carbon-coated in a grid pattern (using finder grid masks; Leica microsystems, #16770162) to locate the region of interest by electron microscopy. Following 1 mg/mL poly-L-lysine (Sigma #P2636) coating overnight, neurons were seeded at densities of 2.5×102 cell/mm2 in 12-well tissue-culture treated

plates (Corning #3512) and cultured in NM0 medium (Neurobasal medium (Gibco #21103049) supplemented with 2% B27 plus (Life Technologies # A3582801) and 1% GlutaMAX (Thermo

Fischer #35050061). STAM-Halo, HRS-Halo and vGlut1-pHluorin were lentivirally transduced at 6 DIV, and experiments performed at 14-15 DIV. To induce activity, 40 µM bicuculine (Bic)

(Tocris #0130/50) and 50 µM 4-aminopyridine (4AP) (Tocris #940) were added to the culture media at 37C. Following 2 h incubation, half of the media was changed to NM0 with 2 mM

Jenelia-Halo-549 nm (Lavis Lab #SB-Jenelia-Halo-549) and incubated for an additional 30 min at 37C. Neurons on the sapphire disks were washed 3X with PBS, fixed with 4% paraformaldehyde (EMS #15714) and 4% sucrose (Sigma #S0389) in PBS for 20 min at room temperature, washed again 3X with PBS, and mounted on a confocal scanning microscope (Carl

Zeiss; LSM880) using a chamber (ALA science #MS-508S). Z-series of fluorescence images were acquired using a 40x objective lens at 2048x2048 pixel resolution. Differential interface contrast images of the carbon grid pattern and neurons on the sapphire disks were also acquired to locate the cells in later steps.

Following fluorescence imaging, neurons were prepared for electron microscopy. Here, sapphire disks were placed into 12-well plate containing 1 mL of 2% glutaraldehyde (EMS

#16530) and 1 mM CaCl2 (Sigma-Aldrich # 63535) in 0.1 M sodium cacodylate (Sigma-Aldrich

#C0250) pH 7.4 and incubated for 1 h on ice. Subsequently, the sapphire disks were washed with

1 mL of chilled 1 M cacodylate buffer 3X for 5 min each. Neurons were then post-fixed for 1 h on ice with 2 mL of 1% osmium tetroxide (EMS #RT19134) and 1% potassium ferrocyanide

(Sigma-Aldrich #P3289) in 0.1 M sodium cacodylate, pH 7.4. After washing 3X with 2 mL of water for 5 min, neurons were stained with 1 mL of 2% uranyl acetate (Polysciences #21447-25) in water for 30 min at room temperature and then briefly washed with 50% ethanol (Pharmco

#111000200) before dehydration in a graded ethanol series (50%, 70%, 90% for 5 min each, and

100% for 5 min 3x). After dehydration, sapphire disks were embedded into 100% epon-araldite resin (Ted Pella #18005, 18060, 18022) with the following schedule and conditions: overnight at

4 ºC, 6 h at room temperature (solution exchange every 2 h), and 48 h at 60 ºC.

For electron microscopy imaging, regions of interest were located based on the bright field images. When a sapphire disk is removed, neurons and carbon-coating remain in the plastic block, which was trimmed to the regions of interest and sectioned with a diamond knife using an ultramicrotome (Leica; UC7). Approximately 40 consecutive sections (80 nm each) were collected onto the pioloform-coated grids and imaged on a Phillips transmission electron microscope (CM120) equipped with a digital camera (AMT; XR80). The fluorescence and electron micrographs were roughly aligned based on size of the pixels and magnification, and the carbon-coated grid patterns. The alignment was slightly adjusted based on the visible morphological features.

Immunofluorescence Microscopy

Neurons or N2a cells were immunostained as described previously (Leal-Ortiz et al.

2008; Sheehan et al. 2016). Briefly, coverslips were fixed with Lorene’s Fix (60 mM PIPES, 25 mM HEPES, 10 mM EGTA, 2 mM MgCl2, 0.12 M sucrose, 4% formaldehyde) for 15 min, incubated with primary and either Alexa-488- or Alexa-647-conjugated secondary antibodies diluted in blocking buffer (2% glycine, 2% BSA, 0.2% gelatin, 50 mM NH4Cl in PBS) for 1 h at room temperature, and washed 4X with PBS between antibody incubations. Coverslips were mounted with DAPI VectaShield (Vector Laboratories) and sealed with clear nail polish.

Alternatively, coverslips were mounted with Aqua-Poly/Mount (Polysciences) and dried

overnight in the dark. Images were acquired with a 40x objective (Neofluar, NA 1.3) or 63x objective (Neofluar, NA 1.4) on an epifluorescence microscope (Axio Observer Z1, Zeiss) with

Colibri LED light source, EMCCD camera (Hamamatsu) and Zen 2012 (blue edition) software.

Images were obtained and processed using Zen Blue 2.1 software. Image processing was done using FIJI (Image-J) software.

In vitro KIF13A/KIF13B transport assay

N2a cells were transfected at 50-70% confluence with BicD2-GFP-FKBP, KIF13A-FRB-

Myc or KIF13B-FRB-Myc, and either mCherry, mCh-Rab5, RFP-Hrs, or STAM-mCh, using

Lipofectamine 3000 (ThermoFisher/Invitrogen) according to manufacturer’s instructions. After

48 h, rapamycin analogue (AP21967) or EtOH control was added to relevant samples as described in (Bentley and Banker 2015). After 3 h of treatment, cells were fixed and imaged as described in Immunofluorescence Microscopy.

Live Imaging

Neurons were imaged at 12-15 DIV after being transduced on 7 DIV, or transfected at 9

DIV. Imaging was conducted in pH-maintaining Air Media, consisting of 250 mL L15 Media

(ThermoFisher/Invitrogen) and 225 mL Hanks Balanced Salt Solution (Sigma), supplemented with 55 mM HEPES, 2% B27 (ThermoFisher), 0.5mM GlutaMAX (ThermoFisher), 0.1%

Pen/Strep, 0.5% Glucose, and 2.5% BME. Neurons were imaged at 37°C with a 40x oil- immersion objective (Neofluar, NA 1.3) on an epifluorescence microscope (Axio Observer Z1,

Zeiss) with Colibri LED light source, EMCCD camera (Hamamatsu) and Zen 2012 (blue edition)

software. One frame was taken every 5 seconds, for a total of 50 frames. Images were obtained and processed using Zen Blue 2.1 software. Axons were identified via location and morphological criteria. Image processing was done using FIJI (Image-J) software.

Live Imaging Analysis

Motility of axonal puncta was analyzed using FIJI’s “Manual Tracker” plugin. Raw data for the movement of each puncta over each video frame were imported into Matlab and motility, directionality, and speed were calculated using a custom program. Only puncta ≥0.5 microns and

<1.5 microns in diameter were analyzed. Directionality of each puncta was categorized as follows: ≥4 µm away from the cell body = anterograde, ≥4 µm towards the cell body = retrograde, ≥4 µm total, but <4 µm in one direction = bidirectional, <4 µm = stationary. Axon length was measured with FIJI software.

Immunoprecipitation and Western Blot

For immunoprecipitation assays, N2a cells were transfected at 50-70% confluence using

Lipofectamine 3000 (ThermoFisher/Invitrogen) according to manufacturer’s instructions. Cells were collected 48 h later in lysis buffer (50 mm Tris-Base, 150 mm NaCl, 1% Triton X-100,

0.5% deoxycholic acid) with protease inhibitor mixture (Roche) and clarified by centrifugation at high speed (20,000 x g). Resulting supernatant was incubated with Dynabeads

(ThermoFisher/Invitrogen) coupled to anti-mCherry antibodies (polyclonal; Biovision). Lysates and beads were incubated at 4°C under constant rotation for 2 h. Beads were washed 2–3X with

PBS containing 0.05% Triton (PBST) and then once with PBS. Bound proteins were eluted using

sample buffer (Bio-Rad) and subject to SDS-PAGE immunoblotting as described following. For

High K+ stimulation of N2a cells, cells were incubated with a High K+ Tyrodes Solution (31.5 mM NaCl, 50 mM KCl, 2 nM MgCl2, 2 mM CaCl, 25 mM HEPES pH 7.4, 30 mM glucose), or a

Control Tyrodes (119 mM NaCl, 2.5 mM KCl, 2 nM MgCl2, 2 mM CaCl, 25 mM HEPES pH

7.4, 30 mM glucose) for 2 h at 37°C before collection in lysis buffer. For all other immunoblotting experiments, neurons were collected directly in 2X SDS sample buffer (Bio-

Rad). Samples were subjected to SDS-PAGE, transferred to nitrocellulose membranes, and probed with primary antibody in 5% BSA/PBS + 0.05% Tween 20 overnight at 4°C, followed by

DyLight 680 or 800 anti-rabbit, anti-mouse, or anti-goat secondary antibodies (ThermoFisher) for 1 h. Membranes were imaged using an Odyssey Infrared Imager (model 9120, LI-COR

Biosciences). Protein intensity was measured using the “Gels” function in FIJI software.

Statistical Analyses

All statistical analyses were performed using GraphPad Prism software. Statistics for each experiment are described in the Figure Legends. A D’Agostino-Pearson normality test was used to assess normality. An unpaired Student’s t test or a Mann–Whitney U test was used to compared two datasets, and an ANOVA or a Kruskal-Wallis test was used to compare three or more datasets. For multiple comparisons, Dunnett’s or Dunn’s post-hoc test was used to compare a control mean to other means, while Sidak’s post-hoc test was used to compare pairs of means.

Statistical significance is noted as follows (and in Figure Legends): ns, P>0.05; *, P≤0.05; **,

P≤0.01; ***, P≤0.001. All experiments were repeated with at least 3 independent sets of samples, except where otherwise noted.

Chapter 3: ESCRT-0 proteins Hrs and STAM are bidirectionally motile in axons, and uniquely regulated by neuronal firing

Abstract

Previous work (Sheehan et al. 2016) suggests that high levels of neuronal firing increase the axonal and synaptic localization of ESCRT proteins Hrs and CHMP2B (the mammalian homologue of ESCRT-III protein VPS2). This suggests that ESCRT proteins undergo activity- dependent changes in their motility and transport to synapses. As ESCRT-0 is responsible for the recruitment of the rest of the ESCRT pathway to specific endosomal sites, we hypothesized that this component would show the most robust response to neuronal firing. ESCRT-0 protein Hrs is the first component of the ESCRT pathway, responsible for sequestering ubiquitinated cargo and recruiting subsequent ESCRT machinery, steps that necessarily precede endolysosomal degradation of SV and other membrane proteins. Thus, we first examined the axonal transport dynamics of Hrs, followed by its binding partner STAM1. We found that the axonal motility of both ESCRT-0 proteins was highly responsive to neuronal firing. To determine whether the transport of other components of the ESCRT pathway is sensitive to neuronal activity, we next examined the axonal dynamics of ESCRT-I protein TSG101 and small Rab GTPase Rab35 at baseline and in response to activity. TSG101 interacts with both Hrs and ubiquitinated cargo on the endosomal membrane, while Rab35 binds Hrs in response to activity, and is involved in

ESCRT-mediated degradation of synaptic vesicle proteins (Sheehan et al. 2016). Of these proteins, ESCRT-0 components alone are sensitive to high levels of neuronal activity, suggesting the potentially exclusive regulation of ECRT-0 motility by neuronal firing.

Results

Hrs and STAM are bidirectionally motile at baseline, and exhibit increased motility in response to neuronal firing

To facilitate analysis of the directional movement of Hrs and other proteins in axons, we utilized tripartite microfluidic chambers that isolate neuronal cell bodies from axons while enabling synapse formation between neurons in adjacent compartments (Figure 5). Mature synapses are able to form in cell body compartments, but do not form in the axonal microgrooves because dendrites cannot enter. Lentiviral transduction of cell bodies in these chambers led to moderate protein expression levels and consistent transduction efficiency across constructs used in this study (~50%; Figure 6A). Hippocampal neurons were transduced on 7 DIV with RFP-Hrs, and live imaging of axons was performed on 13-15 DIV. Axons were imaged in a window that extended at the minimum from 0 to 200 microns from the soma, dependent on the location of the cell body, which was not recorded. Imaging was performed either immediately following application of DMSO (control) or bicuculline and 4-aminopyridine (acute Bic/4AP), or after 2 hours of treatment with these drugs (2h Bic/4AP)((Sheehan et al. 2016); Figure 7A-C). Bic/4AP treatments did not cause any detectable changes in neuronal health or increase cell death over this timeframe (Figure 6B). Under control conditions, we observed that >60% of Hrs puncta were stationary over the course of the ~4-minute imaging session (Figure 7A, D). Acute treatment with Bic/4AP led to a slight but significant increase in puncta motility (from ~38% to

~50% motile), with the effect growing stronger after 2h of treatment (~60% motile puncta;

Figure 7B-D). No change in speed was observed across the three conditions (Figure 7E), nor was there an increase in total axonal Hrs (Figure 7G) indicating that activity leads to the recruitment of Hrs puncta into the motile pool rather than to a fundamental change in the mechanism of

transport. Interestingly, dampening neuronal activity with 2 h TTX treatment, which prevents neuronal firing, did not conversely decrease Hrs motility (Figure 6C), indicating that basal activity levels in these cultures are low. A breakdown of the directionality of Hrs puncta shows that they exhibit a mixture of anterograde, retrograde, and non-processive bidirectional motility

(Figure 7F). However, we found that only anterograde and bidirectional motility increase following treatment with Bic/4AP (Figure 7F), suggesting that neuronal activity preferentially mobilizes Hrs towards distal axonal and presynaptic sites.

Figure 6. Characterization of cell health, lentiviral transduction efficiency, and neuronal activity effects in microfluidic chambers. (A) Lentiviral transduction efficiency of neurons in microfluidic chambers. Graph shows a similar percentage of cell bodies transduced for each lentivirus (≥3 separate experiments, n=28 (Hrs), 23 (STAM), 26 (Rab5), 29, (Rab35)). (B) Analysis of neuronal health in chambers. Graph shows percentage of cells that exhibit NucRed Dead fluorescence following treatments, which is similar across conditions (P>0.05, one-way ANOVA with Dunnett's multiple comparisons test; ≥3 separate experiments, n=33 (control), 24 (acute activity), 31 (2 h activity) videos/condition). (C) Percentage of axonal Hrs puncta that exhibit motility under control conditions, 2 h Bic/4AP treatment, and 2 h TTX treatment to block neuronal activity. TTX treatment does not alter puncta motility compared to the control condition (*P<0.05, one-way ANOVA with Dunnett's multiple comparisons test; ≥3 separate experiments, n=10 (control), 15 (2 h Bic/4AP), 1 (2 h TTX) videos/condition).

We next characterized the axonal motility of STAM1, which is recruited to endosomal membranes by Hrs to form the heterotetrameric ESCRT-0 complex (Mayers et al. 2011;

Takahashi et al. 2015). We found that STAM1, like Hrs, has a punctate pattern of expression in

axons, with similar dynamics to Hrs at baseline (~40% of puncta motile; Figure 7H) and an even larger increase in motility after 2 h of Bic/4AP (~80% motile; Figure 7J). Also similar to Hrs,

Bic/4AP treatment did not alter the average speed of motile STAM1 puncta, which remained

~0.3 m/second (Figure 7L) and yielded only a slight increase in the total number of axonal

STAM1 puncta after 2 h of Bic/4AP (Figure 7N). The increase in the fraction of anterogradely moving puncta was even more pronounced for STAM1 than for Hrs and was accompanied by an increase in retrograde puncta movement (Figure 7M). These findings indicate that the motility of the entire ESCRT-0 complex is sensitive to neuronal firing.

Figure 7. Neuronal activity stimulates the transport of ESCRT-0+ vesicles in axons. (A-C) Representative images and corresponding kymographs of microfluidically isolated axons from 13-15 DIV hippocampal neurons transduced with RFP-Hrs, under control conditions (A), acute treatment with Bicuculline/4AP (B), and 2 h treatment with Bicuculline/4AP (C). Scale bar, 10 µm. (D) Percentage of motile Hrs puncta in axons, showing a significant increase with acute and 2 h Bic/4AP treatment (****P<0.0001, *P<0.05; one-way ANOVA with Dunnett's multiple comparisons test; ≥3 separate experiments, n=20 (control), 15 (acute Bic/4AP), 16 (2 h Bic/4AP) videos/condition). (E) Average speed of motile Hrs puncta (µm/sec), showing no change across conditions. Same ‘n’ values as D. (F) Breakdown of directional movement of Hrs puncta, expressed as percentage of total Hrs puncta. Anterograde and bidirectional movement are significantly increased following 2 h Bic/4AP treatment (****P<0.0001, ***P<0.001, **P<0.01, one-way ANOVA with Dunnett's multiple comparisons test; same ‘n’ values as D). (G) Total number of Hrs-positive puncta in axons, showing no change across conditions. Same ‘n’ values as D. (H-N) Representative images and corresponding kymographs of microfluidically isolated axons from 13-15 DIV hippocampal neurons transduced with STAM1-mCh, under control (H), acute Bic/4AP (I), and 2 h Bic/4AP (J) conditions. Scale bar, 10 µm. (K) Percentage of motile STAM1 puncta in axons, showing a significant increase with acute and 2 h Bic/4AP treatment (****P<0.0001,**P<0.01, *P<0.05, one-way ANOVA with Dunnett's multiple comparisons test; ≥3 separate experiments, n=32 (control), 36 (acute Bic/4AP), 13 (2 h Bic/4AP) videos/condition). (L) Average speed of motile STAM1 puncta (µm/sec), showing no change across conditions. Same ‘n’ values as K. (M) Breakdown of directional movement of STAM1 puncta, showing that anterograde, retrograde, and bidirectional movement are all significantly increased following 2 h Bic/4AP treatment (****P<0.0001, ***P<0.001, **P<0.01, *P<0.05, one-way ANOVA with Dunnett's multiple comparisons test; same ‘n’ values as K). (N) Total number of STAM1-positive puncta in axons, showing only a small change after 2 h Bic/4AP (*P<0.05, one-way ANOVA with Dunnett's multiple comparisons test; same ‘n’ values as K).

ESCRT-I component TSG101 and small GTPase Rab35 do not show activity-dependent motility changes

Using the microfluidic system described above, we then characterized the axonal motility of EGFP-TSG101. We found that TSG101 looked significantly less punctate in neurons than

ESCRT-0 proteins. This distribution may indicate that ESCRT-I does not stably associate with the membrane in the absence of ESCRT-0, which we were not concurrently overexpressing

(Figure 8G-I). However, there were enough TSG101 puncta in the axon to enable their tracking and analysis. We found that TSG101 had similar baseline levels of activity to ESCRT-0 proteins

(~40% of puncta motile; Figure 8A) but did not show an activity-dependent change in motility, speed, or number of puncta per 100 µm axon (Figure 8A-F). Indeed, the only difference in

TSG101 motility between control and high-activity conditions was a slight increase in

bidirectional motility after 2 h of Bic/4AP treatment (Figure 8E). This led us to conclude that the motility of TSG101 is either not significantly altered by neuronal activity, or that the effect appears only after more prolonged neuronal firing. The latter possibility could be the purview of future work.

Figure 8. ESCRT-I protein TSG101 shows little change in motility in response to neuronal activity. (A-B) Representative images and corresponding kymographs of microfluidically isolated axons from 13-15 DIV hippocampal neurons transduced with EGFP-TSG101, under control conditions (A) and after 2 h treatment with Bicuculline/4AP (B). Scale bar, 10 µm. (C) Percentage of motile TCG101+ puncta in axons, showing no significant increase with acute or 2 h Bic/4AP treatment (P>0.05; one-way ANOVA with Dunnett's multiple comparisons test; ≥3 separate experiments, n=30

(control), 33 (acute Bic/4AP), 14 (2 h Bic/4AP) videos/condition). (D) Average speed of motile TSG101+ puncta (µm/sec), showing no change across conditions. Same ‘n’ values as C. (E) Breakdown of directional movement of TSG101 puncta, expressed as percentage of total TSG101+ puncta. Only bidirectional movement is significantly increased following 2 h of Bic/4AP treatment (*P<0.05, one- way ANOVA with Dunnett's multiple comparisons test; same ‘n’ values as C). (F) Total number of TSG101+ puncta in axons, showing no change across conditions. Same ‘n’ values as C. (G-H) Representative images from neurons 12 DIV co-transfected with RFP-Hrs (G) and EGFP-TSG101 (H). Arrows show colocalized puncta. TSG101 becomes more punctate and exhibits strong colocalization with Hrs when they are co-transfected. (I) Representative images of EGFP-TSG101 in the absence of Hrs, showing a more diffuse expression pattern.

Given our previous findings that Rab35 is present on the same vesicles as Hrs and is required for activity-dependent degradation of SV proteins (Sheehan et al. 2016), we also investigated the transport dynamics of axonal Rab35. We found that EGFP-Rab35 exhibited higher motility than Hrs at baseline (~60% motile, Figure 9A), but no change in motility in response to Bic/4AP, even after 2 h of treatment (Figure 9A-G). However, the colocalization between RFP-Hrs and EGFP-Rab35 in axons increased slightly but significantly in response to 2 h Bic/4AP (Figure 9H-I). These data, along with similarities in the directional transport profiles of Hrs and Rab35 following 2 h Bic/4AP treatment (Figure 7F & Figure 9F), suggest that Rab35 colocalizes more strongly with the motile pool of Hrs, and that its motility is not regulated by neuronal firing in the same way.

Figure 9. Neuronal activity does not stimulate Rab35 motility in axons. (A-C) Representative images and corresponding kymographs for 13-15 DIV hippocampal axons transduced with EGFP- Rab35, under control conditions (A), acute treatment with Bicuculline/4AP (B), and 2 h treatment with Bicuculline/4AP (C). Scale bar, 10 µm. (D) Percentage of motile Rab35+ puncta in axons, showing no significant difference across conditions (P>0.05, Kruskal-Wallis test with Dunn's multiple comparisons test; ≥3 separate experiments, n= 20 (control), 15 (acute Bic/4AP), 20 (2 h Bic/4AP) videos/condition). (E) Average speed of motile Rab35+ puncta (µm/sec), showing no change across conditions. Same ‘n’

values as D. (F) Breakdown of directional movement of Rab35+ puncta, expressed as percentage of total puncta, showing no change across conditions. (G) Representative images of fixed hippocampal axons expressing RFP-Hrs and EGFP-Rab35. Scale bar, 10 µm. (H) Percentage of Hrs puncta that are positive for Rab35, showing increased colocalization following 2 h Bic/4AP treatment (*P<0.05, one- way ANOVA with Dunnett's multiple comparisons test; ≥3 separate experiments, n=36 (control), 39 (2 h Bic/4AP) images/condition). All scatter plots show mean ± SEM.

Summary

Altogether, these results suggest that of the ESCRT-0 and -I components and related Rab

GTPase characterized here, only ESCRT-0 proteins Hrs and STAM1 show a strong response to neuronal firing. These findings indicate that the axonal motility of ESCRT-0 proteins may be uniquely regulated by neuronal activity. This study is the first to investigate the motility of

ESCRT components in neurons and has yielded interesting information about the behavior of these proteins in axons. In contrast to the highly processive motility of autophagosomes (0.45

µm/s) (Maday, Wallace, and Holzbaur 2012; Maday and Holzbaur 2016; Maday et al. 2014), we find that ESCRT proteins exhibit more complex bidirectional motility, as well as more stationary behavior in axons (Figure 7F&M, Figure 8E). These findings suggest not only that distinct axonal motor proteins are responsible for the transport of endolysosomal versus autophagic structures, but also that these pathways may have divergent roles at the presynapse. In contrast to the constitutive formation and flow of autophagosomes from presynaptic terminals to the soma, which is unaffected by stimuli such as starvation or neuronal activity (Maday, Wallace, and

Holzbaur 2012; T. Wang et al. 2015), the transport of ESCRT proteins appears to be more heterogeneous and susceptible to regulation. This work further demonstrates the distinctive dynamics and roles of the autophagic and ESCRT degradative machinery in neurons, potentially illuminating the roles of each pathway in the degradation of synaptic vesicle proteins under different conditions (Hernandez et al. 2012; Sheehan et al. 2016; Uytterhoeven et al. 2011, 2017;

Vijayan and Verstreken 2017; Birdsall and Waites 2019). However, more work is needed to fully dissect the substrates and functions of these interconnected pathways.

It is also significant that the activity-dependent changes in motility are observed for

ESCRT-0 but not ESCRT-I proteins. Given the layered recruitment of ESCRT complexes to the endosomal membrane (Katzmann, Babst, and Emr 2001; Henne, Buchkovich, and Emr 2011;

Raiborg, Rusten, and Stenmark 2003; Schuh and Audhya 2014), it is possible that later complexes do not need to be responsive to neuronal firing, as recruitment of ESCRT-0 will initiate recruitment of the rest of the pathway. Indeed, many downstream ESCRT proteins cannot stably associate with membranes in the absence of an ESCRT-0 interaction. It is also possible that downstream ESCRT components are responsive to activity, but over a longer time-course than assessed in our experiments. Future work will address this possibility by further characterizing the axonal transport and motility of ESCRT-II and -III proteins.

Finally, the relationship between neuronal activity and Rab35 localization and transport remains murky. Although the motility of Rab35 appears to be unaffected by neuronal firing

(Figure 9), our previous findings that Rab35 and Hrs localize to the same vesicles (Sheehan et al.

2016), as well as the uncanny similarities between the transport profiles of Hrs under high- activity conditions and Rab35 across conditions (Figure 7F,Figure 9F) and the increased colocalization of Hrs with Rab35 in response to activity (Figure 9H-I), all suggest a possible recruitment of Hrs to Rab35+ vesicles in response to activity. This would be consistent with the fact that Hrs is an effector of Rab35, and is recruited by GTP-Rab35 in response to neuronal activity (Sheehan et al. 2016).

Chapter 4: Activity-responsive Hrs+ vesicles are a subset of total Rab5+ vesicles

Rationale

Hrs typically associates with early endosomes (Sun et al. 2003; Raiborg, Rusten, and

Stenmark 2003; Flores-Rodriguez et al. 2015; Raiborg et al. 2001); thus, we examined whether other early endosomal proteins exhibit similar transport dynamics and activity-dependent motility. Along with the ESCRT-0 complex, another key component of early endosomes, critical for their formation and function, is the small GTPase Rab5 (Gorvel et al. 1991; Deinhardt et al.

2006; Poteryaev et al. 2010). We therefore investigated the motility of Rab5 under control and activity conditions. After observing that the Rab5 motility profile and response to activity did not mirror those of ESCRT-0 proteins, we hypothesized that axonal ESCRT-0+ vesicles, characterized here for the first time, could represent specialized ‘pro-degradative’ endosomes. To address this hypothesis, we investigated the colocalization of Hrs with Rab5, as well as the mechanism that targets Hrs to vesicle membranes. We find that axonal Hrs+ vesicles are a subset of Rab5+ vesicles, and that Hrs association with vesicle membranes requires the canonical

PI3P/FYVE interaction and appears to occur in the soma prior to vesicle transport into axons.

Figure 10. Rab5+ vesicles show a modest increase in motility in response to heightened neuronal activity. (A-C) Representative images and corresponding kymographs of microfluidically isolated axons from 13-15 DIV hippocampal neurons transduced with EGFP-Rab5, under control (A), acute Bic/4AP (B), and 2 h Bic/4AP (C) conditions. Scale bar, 10 µm. (D) Percentage of motile Rab5 puncta in axons, showing a significant increase only after 2 h Bic/4AP treatment (**P<0.01, Kruskal-Wallis test with Dunn's multiple comparisons test; ≥3 separate experiments, n=31 (control), 30 (acute Bic/4AP), 14 (2 h Bic/4AP) videos/condition). (E) Average speed of motile Rab5+ puncta (µm/sec) is unchanged across conditions, as with Hrs and STAM1. Same ‘n’ values as I. (F) Breakdown of directional Rab5+puncta movement, showing that only anterograde movement is significantly increased following 2 h Bic/4AP treatment (**P<0.01, *P<0.05, one-way ANOVA with Dunnett's multiple comparisons test; same ‘n’ values as I). All scatter plots and bar graphs show mean ± SEM. (G) Total number of Rab5+ puncta in axons, showing no change across conditions. Same ‘n’ values as D.

Results

Rab5 shows a modest activity-dependent increase in axonal motility

Utilizing the microfluidic chambers described in Chapter 3, we found that EGFP-Rab5 exhibited higher motility under control conditions than ESCRT-0 proteins (~60% motile puncta at baseline; Figure 10A,D), and that the motility of Rab5 puncta did not increase with acute

Bic/4AP treatment (Figure 10B,D), in contrast to Hrs and STAM1. However, after 2h of

Bic/4AP, Rab5 puncta exhibited significantly increased motility, to ~80% motile puncta (Figure

10C-D). As with Hrs and STAM1, the average speed and number of axonal Rab5 puncta was unaffected by activity treatment (Figure 10E,G). The proportion of anterogradely moving Rab5 puncta also increased significantly after 2 h of Bic/4AP treatment (Figure 10F) and was the only observed change in Rab5 directional movement. These data show that while Rab5 puncta do exhibit activity-induced changes in axonal motility, this effect takes longer to materialize and is less robust than that of activity on ESCRT-0 proteins.

Figure 11. Hrs is present on a subset of Rab5+ vesicles in axons. (A) Representative images and corresponding kymographs for 13-15 DIV hippocampal axons co-transfected with RFP-Hrs and EGFP- Rab5. Arrows show puncta that are positive for both proteins. Scale bar, 10 µm. (B) Percentage of stationary Hrs puncta that are also positive for Rab5, demonstrating no significant difference across conditions (P>0.05, one-way ANOVA with Dunnett's multiple comparisons test; ≥3 separate experiments, n=14 (control), 18 (acute Bic/4AP), 12 (2 h Bic/4AP) videos/condition). (C) Percentage of stationary Rab5+ puncta that are also positive for Hrs, demonstrating no significant differences across condition. Same ‘n’ values as B. (D) Percent of stationary puncta that are positive for both Hrs and Rab5 per 100 µm of axon, showing no significant differences across conditions. Same ‘n’ values as B. (E) Percentage of motile Hrs+ puncta that are positive for Rab5, showing a significant increase in

motility after 2 h of Bic/4AP treatment (*P<0.05, Kruskal-Wallis test with Dunn’s multiple comparisons test; same ‘n’ values as B). (F) Percentage of motile Rab5 puncta that are also positive for Hrs, showing no significant difference in motility across conditions. Same ‘n’ values as B. (G) Percentage of motile puncta that are positive for both Hrs and Rab5 per 100 µm of axon, showing a significance increase in motility after 2 h of Bic/4AP treatment (**P<0.01, one-way ANOVA with Dunnett's multiple comparisons test; same ‘n’ values as B). All scatter plots show mean ± SEM.

Hrs+ vesicles are a subset of Rab5+ vesicles

Since the axonal dynamics of Rab5 do not precisely align with those of ESCRT-0 proteins, we next investigated whether Hrs is present on the same vesicles as Rab5 in axons.

Here, we performed time-lapse imaging of axons from dissociated neurons co-transfected with

RFP-Hrs and EGFP-Rab5 (Figure 11A). The density of Rab5+ puncta in co-transfected axons was on average higher than that of Hrs+ puncta, and while >80% of Hrs+ puncta (stationary and motile) were also positive for Rab5 (Figure 11B,E), the converse was not true. Indeed, only

~40% of stationary Rab5+ puncta and ~60% of motile Rab5 puncta were positive for Hrs (Figure

11C,F). These data suggest that Hrs is present on a subset of Rab5+ structures, consistent with previous studies showing that Rab5 is associated with a heterogenous population of endosomes as well as SVs (Simonsen et al. 1998; Christoforidis et al. 1999; Flores-Rodriguez et al. 2015;

Fischer von Mollard et al. 1994; Shimizu, Kawamura, and Ozaki 2003; K. Zhang et al. 2013).

Moreover, while the total number (stationary and motile) of Hrs+/Rab5+ puncta did not change in response to neuronal activity (Figure 11D), the number of motile Hrs+/Rab5+ puncta increased significantly (Figure 11G). These findings indicate that an Hrs+ subset of Rab5+ vesicles is responsive to neuronal firing, providing a potential explanation for why the total population of axonal Rab5 shows only a modest increase in motility in response to Bic/4AP treatment (Figure

10).

Figure 12. Hrs 2xFYVE domain does not recapitulate the motile behavior of full-length Hrs. (A- C) Representative images and corresponding kymographs for 13-15 DIV hippocampal axons transduced with EGFP-2xFYVE domain, under control conditions (A), acute treatment with Bic/4AP (B), and 2 h treatment with Bic/4AP (C). Scale bar, 10 µm. (D) Percentage of motile 2xFYVE puncta in axons, showing no significant difference across conditions (P>0.05, one-way ANOVA with Dunnett's multiple comparisons test; ≥3 separate experiments, n=11 (control), 12 (acute Bic/4AP), 10 (2 h Bic/4AP) videos/condition). (E) Average speed of motile 2xFYVE+ puncta (µm/sec), showing no change across conditions. Same ‘n’ values as D. (F) Breakdown of directional movement of 2xFYVE+ puncta, expressed as percentage of total puncta, showing no change across conditions. All scatter plots and graphs show mean ± SEM. (G) Total number of 2xFYVE+ puncta in axons, showing no change across conditions. Same ‘n’ values as D.

Figure 13. Hrs vesicle association requires FYVE domain/PI3P interactions and occurs in the somatodendritic compartment. (A-C) Representative images of EGFP-Hrs localization in cell bodies and dendrites of 13-15 DIV hippocampal neurons under control conditions (A), after 2 h of SAR405 treatment (B), and with the R183A FYVE domain inactivating mutation (C). (D) Number of Hrs puncta/µm2 in the somatodendritic compartment under these conditions, demonstrating significant decreases in puncta density with SAR405 treatment or R183A mutation (***P<0.001, *P<0.05, one- way ANOVA with Dunnett's multiple comparisons test; ≥3 separate experiments, n=33 (control), 35 (2 h SAR405), 17 (FYVE mutant) images/condition). (E-F) Representative images and corresponding kymographs of microfluidically isolated hippocampal axons transduced with EGFP-Hrs, under control conditions (E) and after 2 h of treatment with SAR405 (F). (G) Number of total Hrs puncta per 100 µm of axon, showing no significant difference across these conditions (P>0.05, unpaired t-test; ≥3 separate experiments, n=11 (control), 10 (2 h SAR405) videos/condition). (H) Average speed of motile Hrs puncta (µm/sec), showing no difference following SAR405 treatment. Same ‘n’ values as B. (I) Percentage of motile Hrs puncta in axons, showing no difference following SAR405 treatment. Same ‘n’ values as B. All scatter plots show mean ± SEM.

Hrs targeting to vesicle membranes depends on the FYVE domain/PI3P interaction and occurs in the soma

Studies in non-neuronal cells show that Hrs recruitment to endosomal membranes depends on interactions between its FYVE domain and the endosome-enriched lipid phosphoinositol-3-phosphate (PI3P) (Komada & Soriano, 1999; Stahelin et al., 2002). However, other domains of Hrs, in particular the C-terminal coiled-coil domain, have also been reported to facilitate its membrane targeting (Hayakawa & Kitamura, 2000; Raiborg et al., 2001). Given the dearth of evidence about axonal targeting of Hrs, and reports that EEA1, another FYVE domain early endosomal protein, is not present in axons (Gaullier et al., 2000; Lawe et al., 2000; Wilson et al., 2000), we wanted to investigate whether the FYVE/PI3P interaction was required for vesicular targeting of axonal Hrs. We first assessed whether the axonal dynamics of the Hrs

FYVE domain replicated those of full-length Hrs. Here, neurons in microfluidic chambers were lentivirally transduced with EGFP-2xFYVE, and their dynamic behavior assessed under control, acute Bic/4AP, and 2 h Bic/4AP conditions. Interestingly, while the 2xFYVE domain exhibited highly punctate expression in axons, its dynamics were different from those of full-length Hrs and it did not demonstrate activity-dependent changes in motility (Figure 12A-G), indicating that

other domains of Hrs facilitate its activity-dependent transport. We next assessed whether the

FYVE/PI3P interaction was necessary for targeting Hrs to vesicle membranes in neurons. For these experiments, we treated EGFP-Hrs-expressing neurons with an inhibitor of the phosphatidylinositol 3-kinase VPS34, required for synthesis of PI3P (SAR405; (Schu et al.

1993)). Two hours of treatment with the inhibitor dramatically decreased the number of somatodendritic Hrs puncta compared to vehicle control (Figure 13A-B,D). This effect was replicated by an inactivating mutation of the Hrs FYVE domain (R183A; (Raiborg et al. 2001))

(Figure 13C-D), indicating that the FYVE/PI3P interaction is critical for membrane targeting of

Hrs in neurons. However, 2 h treatment with SAR405 did not alter the number or motility of axonal Hrs puncta (Figure 13E-I), suggesting that Hrs association with vesicles occurs in the somatodendritic compartment prior to axonal transport, and remains stable in axons over the course of several hours.

Summary

Although studies in yeast and other eukaryotic cells have firmly established the presence of ESCRT-0 proteins on early endosomes (Sun et al. 2003; Raiborg, Rusten, and Stenmark 2003;

Flores-Rodriguez et al. 2015; Raiborg et al. 2001; Henne, Buchkovich, and Emr 2011), the localization of ESCRT-0 proteins in axons have never been described. Moreover, it is well known that neurons contain a multitude of distinct endosomal subtypes not seen in other cells, which are often localized to a specific subcellular compartment (Yap et al. 2018; Maday and

Holzbaur 2016; Kulkarni and Maday 2018b, 2018a). To better understand the function and regulation of the ESCRT pathway in neurons, we investigated the axonal pool of ESCRT-0 structures. Our findings suggest that axonal Hrs+ vesicles are a subset of total Rab5+ vesicles,

representing by our count ~40-60% of the total pool of axonal Rab5+ structures (Figure 11C,F).

These ‘pro-degradative’ vesicles are likely distinct from other subsets of Rab5 endosomes previously identified (Olenick, Dominguez, and Holzbaur 2018; Goto-Silva et al. 2019), and based on our data (Figures 10&11) it seems that only this subset of axonal Rab5 vesicles is sensitive to neuronal activity. We hypothesize that these Hrs+/Rab5+ structures represent either early endosomes destined to become MVBs via later ESCRT protein recruitment, or transport vesicles bringing degradative machinery to the presynapse. Future studies will further explore these possibilities.

It is also remarkable that while the FYVE domain clearly regulates Hrs association with membranes (Figure 13), it does not regulate the activity-induced motility of Hrs in axons

(Figures 7&12), suggesting that other domains are responsible for Hrs interaction with motor proteins. Moreover, unlike somatodendritic Hrs, axonal Hrs does not dissociate from vesicles within a 2-hour timeframe in response to a reduction in PI3P levels (Figure 13E-I). These findings provide the first indication that the vesicle association of the axonal pool of Hrs is different from that of the somatodendritic pool. The mostly likely interpretation of these data is that Hrs is assembled onto vesicles before their transport into axons, and that Hrs association with membranes in axons is more stable than that in the somatodendritic compartment. The apparent stability of axonal Hrs on vesicle membranes also suggests that these structures are not formed in response to neuronal activity, but rather that existing Hrs+ structures are mobilized by activity, as least in the 2 h time course that we examined. Future work will identify the factors that create this heightened stability. Overall, these data begin to elucidate the identity of axonal

ESCRT-0 transport vesicles and brings us a step closer to understanding how the ESCRT pathway is dynamically regulated in neurons.

Chapter 5: Hrs-positive vesicles interact with SV pools

Rationale

Chapters 3 and 4 demonstrate that ESCRT-0+ vesicles are more motile in response to neuronal activity and comprise a subset of Rab5-positive structures. Previous work from our lab shows that the ESCRT pathway mediates the activity-dependent degradation of SV proteins, and that neuronal firing increases the localization of ESCRT proteins to axons as well as their colocalization with SV pools (Sheehan et al. 2016). Thus, we hypothesized that an increase in neuronal activity would stimulate the recruitment of ESCRT-0+ vesicles to presynaptic boutons.

To test this, we studied the colocalization of motile Hrs puncta with synaptic vesicle markers, in order to assess whether moving Hrs endosomes could be seen ‘pausing’ at presynaptic sites. We found that such pausing does occur, but its duration is not influenced by neuronal activity. Since we identified Hrs+ structures as a subset of Rab5+ vesicles, we next assessed whether these structures exhibited typical endosomal morphology, and whether neuronal firing causes an increase in the number of endosomal structures within presynaptic boutons. We find that Hrs+ structures in axons appear to be degradative endosomes, and that there are more endosomes and

MVBs at synapses in higher activity conditions.

Figure 14. Motile Hrs interacts with presynaptic sites in a ubiquitin-dependent mechanism. (A) Representative images and corresponding kymographs for hippocampal axons expressing RFP-Hrs and EGFP-Synapsin. Arrows show Hrs+ puncta that colocalize with Synapsin. Scale bar, 10 µm. All scatter plots show mean ± SEM. (B) Immunoblot of lysates from HEK-293 cells transfected with VAMP2 and treated for 2 h with DMSO or deubiquitinating enzyme inhibitor PR619, then immunoprecipitated (IP) with VAMP2 antibody and probed with ubiquitin antibodies. Note that PR619 increases VAMP2 ubiquitination. (C) Percentage of time that motile Hrs puncta are colocalized with Synapsin puncta during the 4 min imaging session, which does not change across conditions (P>0.05, one-way ANOVA with Dunnett's multiple comparisons test; ≥3 separate experiments, n=18 (control), 28 (acute Bic/4AP), 28 (2 h Bic/4AP) videos/condition). (D) Percentage of time that that motile Hrs puncta colocalize with Synapsin puncta after treatment with Bic/4AP +/- PR619 for 2 h, showing a significant increase in the presence of PR619 (*P<0.05, unpaired t-test; ≥3 separate experiments, n=19 (Bic/4AP), 7 (Bic/4AP+PR619) videos/condition). All scatter plots show mean ± SEM.

Results

Motile Hrs puncta colocalize with stationary Synapsin puncta in the axon

Like other early endosome-associated proteins (e.g. Rab5) (Goto-Silva et al. 2019), Hrs exhibits a mixture of anterograde, retrograde, and bidirectional movement in axons (Figure 7F).

However, treatment with Bic/4AP increases only the anterograde and bidirectional motility of

Hrs (Figure 7F), signifying that neuronal activity preferentially increases Hrs transport towards more distal presynaptic terminals. To investigate whether this mobilization leads to increased

Hrs contact with SV pools, we co-transfected dissociated hippocampal neurons with RFP-Hrs and EGFP-Synapsin (to label SV pools/presynaptic boutons (Leal-Ortiz et al. 2008)), and performed live imaging to assess colocalization between motile Hrs and stationary Synapsin puncta in axons. While we did observe motile Hrs puncta colocalizing with Synapsin at presynaptic sites (Figure 14A), this behavior did not depend upon neuronal activity, as motile

Hrs puncta spent ~35% of their time in contact with Synapsin regardless of activity condition

(Figure 14C). However, Hrs/Synapsin colocalization was significantly increased following 2 h of treatment with PR619 (Figure 14C), a deubiquitinating enzyme inhibitor that promotes SV protein ubiquitination (Figure 14B), suggesting that Hrs interaction with presynaptic sites is ubiquitin-dependent. This finding is consistent with the role of Hrs in binding ubiquitinated proteins and sorting them toward MVBs.

CLEM reveals that axonal Hrs-positive vesicles are degradative structures

To examine the morphology of Hrs-positive vesicles in axons and at presynaptic boutons, we performed correlative-light electron microscopy on cultured neurons lentivirally co-

transduced with vesicular glutamate transporter (vGlut1-GFP) to label SV pools and HaloTag-

Hrs. Interestingly, Hrs-positive structures identified in the axon appear to be degradative in nature, including an endosome and a multivesicular body (Figure 15A). These data, along with our previous findings (Sheehan et al. 2016), suggest that Hrs is associated with degradative structures in axons.

Figure 15. Correlative light electron microscopy (CLEM) reveals that Hrs+ vesicles can be degradative structures. (A) Representative CLEM images of axons from hippocampal neurons transduced with vGlut1-EGFP (green) and Halo-Hrs (+ Jenelia-Halo-549 nm; red), treated for 2 h with

Bic/4AP. Each set of images has a lower magnification image of the putative axon, and a higher magnification image of the Halo-Hrs-positive structure (white arrows).

Endosomal structures are more prevalent at the synapse in high-activity conditions

In our earlier study, we demonstrated that neuronal firing increases the recruitment of

ESCRT proteins into axons and their colocalization with SV markers (Sheehan et al. 2016).

Given the role of ESCRT proteins in catalyzing MVB formation (Raiborg, Rusten, and Stenmark

2003), these changes in ESCRT protein localization may increase numbers of MVBs and other endosomal structures at presynaptic boutons. We therefore examined electron micrographs of boutons from 16-21 DIV hippocampal neurons treated for 48 h with tetrodotoxin (TTX) to reduce neuronal firing or Gabazine to increase firing. We observed a significantly higher density of presynaptic MVBs and endosomes (defined as single-membrane structures >80 nm in diameter) following gabazine versus TTX treatment (Figure 16A,C), suggesting that neuronal activity promotes the appearance of these structures at presynaptic boutons. Since other studies have reported that neuronal firing stimulates the biogenesis of autophagosomes at presynaptic terminals (T. Wang et al. 2015), we also counted the number of multilamellar structures present under each condition. Interestingly, not only was the average density of these structures lower than that of endosomes/MVBs regardless of activity level (~0.1 multilamellar structures/m2 vs.

~1 endosomes/m2 at baseline), but we also did not see an increase with Gabazine treatment

(Figure 16B,D), indicating that prolonged neuronal activity specifically increases the number of endosomal structures and MVBs at presynaptic boutons.

Figure 16. Neuronal activity increases the number of presynaptic endosomes. (A-B) Representative electron micrographs of endosomal structures (single-membrane, >80nm diameter, red arrows; A) and multilamellar structures (>1 membrane, red arrows; B) present at presynaptic boutons from 16-21 DIV hippocampal neurons treated for 48 h with TTX or gabazine. Scale bar, 500 nm. (C- D) Quantification of endosomal structures (C) and multilamellar structures (D) per µm2 of presynaptic area at TTX- or gabazine-treated boutons. The density of endosomal structures, but not multilamellar structures, is significantly increased following gabazine treatment. Scatter plot shows median and interquartile range (**** P<0.0001, Mann-Whitney test; n=102 (TTX-treated) and 106 (gabazine- treated) boutons/condition).

Summary

The role of the endolysosomal pathway in degrading specific SV proteins (Sheehan et al.

2016; Uytterhoeven et al. 2011, 2017), in concert with the findings that ESCRT proteins are more motile in the anterograde direction (Figure 7) and exhibit increased colocalization with presynaptic markers after prolonged neuronal firing (Sheehan et al. 2016), strongly suggest that

ESCRT-0 proteins are recruited to the synapse by neuronal activity in order to degrade SV proteins. In this chapter, we provide more definitive evidence for this model, showing that motile

Hrs+ vesicles spend ~35% of their time in contact with presynaptic markers, and that this percentage can be increased by enhancing protein ubiquitination. Neuronal activity also increases the number of anterogradely and bidirectionally motile Hrs puncta (Figure 7), thereby increasing the likelihood that Hrs vesicles interact with SV pools. We also analyzed the morphology of Hrs- positive structures in axons and at presynapses using correlative light electron microscopy

(CLEM). Although we could not obtain a large enough sample for statistical analysis, we found that axonal structures carrying Hrs exhibited morphological hallmarks of degradative endosomes and MVBs (Figure 15A). Finally, we determined statistically that neuronal activity increases the number of endosomes within presynaptic boutons (Figure 16A,C), providing evidence that neuronal activity increases the formation and/or delivery of endosomes to presynapses.

Altogether, these data support the hypothesis that ESCRT-0 endosomes are transported to SV pools in response to neuronal activity, in order to initiate the degradation of synaptic vesicle proteins. Future experiments will yield further insights into the regulation of the endolysosomal system at the presynapse.

Chapter 6: KIF13A is responsible for activity-dependent anterograde movement of Hrs in the axon, and necessary for the turnover of ESCRT- degraded SV proteins

Rationale

We have shown that axonal Hrs+ puncta exhibit pronounced changes in directional motility following 2 h Bic/4AP treatment, in particular a >2-fold increase in anterograde transport (Figure 7F). These findings suggest that a plus-end directed kinesin motor protein mediates at least some element of the activity-dependent transport of Hrs. Multiple kinesins have been implicated in the axonal transport of various cargo (Jenkins et al. 2012; Bentley et al. 2015;

Maday et al. 2014), creating a panoply of potential candidates. However, previous studies have identified kinesins specifically responsible for the plus-end directed transport of Rab5+ early endosomes, including the kinesin-3 family members KIF13A and KIF13B (Bentley et al., 2015,

Journal of Cell Biology; Huckaba et al., 2011, Journal of Biological Chemistry). Since Hrs+ axonal puncta are a subset of Rab5 structures (Figure 11B,E), this provided a convenient starting point, and we investigated the possibility that KIF13A/B are involved in the axonal transport of

Hrs+ endosomes. Using a split kinesin assay (Bentley and Banker 2015), we found that KIF13A, but not KIF13B, interacts with Hrs+ endosomes. We next evaluated the importance of this motor for activity-dependent ESCRT-0 motility, as well as for SV protein degradation. These studies revealed that knockdown of endogenous KIF13A in cultured neurons ablates the activity-induced increase in ESCRT-0 puncta motility and slows the degradation of SV proteins. Overall, these results suggest that the transport of ESCRT-0+ endosomes by KIF13A is vital to the neuronal function of the ESCRT pathway.

Figure 17. KIF13A and KIF13B mediate transport of Rab5+ vesicles in N2a cells. (A-B) Representative images of N2a cells transfected with BicD2-GFP-FKBP (green) and KIF13A-FRB-Myc (blue; A) or KIF13B-FRB-Myc (blue; B), +/- rapamycin analogue to link FKBP/FRB (linker). (C-D)

Representative images of N2a cells transfected with mCh-Rab5 (red), BicD2-GFP-FKBP (green), and KIF13A-FRB-Myc (blue), +/- linker, and corresponding line scan graphs. White arrows show line scan area. (D) showing that addition of linker induces protein colocalization. (E-F) Representative images of N2a cells transfected with mCh-Rab5 (red), BicD2-GFP-FKBP (green), and KIF13B-FRB-Myc (blue), +/- linker, and corresponding line scan graphs. White arrows show line scan area. (F) showing that addition of linker induces protein colocalization.

Results

KIF13A, but not KIF13B, interacts with Hrs+ vesicles

Previous work provided us with two strong candidates for the plus-end directed transport of ESCRT-0+ vesicles: kinesin-3 motors KIF13A and KIF13B (Bentley et al., 2015, Journal of

Cell Biology; Huckaba et al., 2011, Journal of Biological Chemistry). Using an assay developed to evaluate the ability of specific kinesins to transport particular cargoes (Bentley and Banker

2015), which had been previously used to confirm that these motors facilitate Rab5 motility, we examined whether KIF13A and/or KIF13B mediate the transport of Hrs. We first confirmed the efficacy of this assay in Neuro2a (N2a) cells, as previous work had been done in fibroblasts

(Bentley et al. 2015), by transfecting cells with the minus-end directed motor domain of BicD2

(BicD2-GFP-FKBP), together with the cargo binding domain of KIF13A or KIF13B (KIF13A-

FRB-Myc, KIF13B-FRB-Myc). As previously reported, addition of a membrane-permeant rapamycin analog to link the FRB and FKBP domains led to the cotransport of BicD2 and

KIF13A/B to the cell centrosome (Figure 17A-B). Moreover, we were able to recapitulate transport of Rab5 by both KIF13A and KIF13B as previously observed (Figure 17C-F (Bentley et al. 2015)). We next tested the ability of KIF13A and KIF13B to facilitate transport of RFP-

Hrs. Surprisingly, while Hrs was readily transported to the centrosome in the presence of the linker and KIF13A (Figure 18A-B), it was not transported by KIF13B (Figure 18C-D). These findings indicate a specificity of the transport mechanism for Hrs+ vesicles that does not exist for

all Rab5+ vesicles. We further tested the specificity of interactions between Hrs and KIF13A/B by coimmunoprecipitation in N2a cells co-transfected with RFP-Hrs, STAM1-mCh, or mCherry control, together with the KIF13A- or KIF13B-FRB-Myc constructs. Consistent with our transport assays, we found that Hrs clearly precipitated KIF13A, but not KIF13B (Figure 18E-F).

STAM1 also had some ability to precipitate KIF13A, albeit variably, but did not precipitate

KIF13B (Figure 18E-F). We repeated this coimmunoprecipitation assay in the presence of high

K+ stimulation (50 mM KCl for 2 h; (La Montanara et al., 2015)) to depolarize the N2a cells, and preliminary results indicated that Hrs precipitates more KIF13A after 2 h of stimulation. These data suggest that the Hrs/KIF13A interaction may grow stronger in a depolarized environment

(Figure 19A-B), however more work must be done to validate this finding. Overall, these results suggest that Hrs, and potentially STAM1, interact with KIF13A, and identify this motor protein as a candidate for the activity-dependent transport of axonal ESCRT-0 vesicles.

Figure 18. KIF13A, but not KIF13B, mediates transport of Hrs+ vesicles. (A) Representative images of N2a cells co-transfected with RFP-Hrs (red), BicD2-GFP-FKBP (green), and KIF13A-FRB- Myc (blue), +/- rapamycin analogue to link FKBP/FRB (linker). (B) Line scan graphs of the images in panel A, showing that addition of the linker induces colocalization of these proteins as depicted by a single peak in the lower graph. White arrows show line scan area. (C-D) Representative images of N2a

cells co-transfected with RFP-Hrs (red), BicD2-GFP-FKBP (green), and KIF13B-FRB-Myc (blue), +/- linker (C), and corresponding line scan graphs (D) showing that addition of linker does not induce protein colocalization. (E) Immunoblot of lysates from N2a cells expressing mCh, mCh-STAM, or RFP-Hrs, together with KIF13A-FRB-Myc or KIF13B-FRB-Myc, immunoprecipitated (IP) with mCh antibody and probed with Myc or mCh antibodies. (F) Quantitative analysis of KIF13A/B-FRB-Myc band intensity after immunoprecipitation with Hrs or STAM, expressed as intensity normalized to mCh control. Hrs consistently precipitates KIF13A but not KIF13B (**P<0.001, one-way ANOVA with Dunnett's multiple comparisons test; ≥3 separate experiments). Bars show mean ± SEM.

Figure 19. Depolarization increases the association between KIF13A and Hrs. (A) Immunoblot of lysates from N2a cells expressing mCh, mCh-STAM, or RFP-Hrs, together with KIF13A-FRB-Myc, treated either with 50 mM KCl (high K+) or 2.5 mM KCl (control) for 2 h, then immunoprecipitated (IP) with mCh antibody and probed with Myc or mCh antibodies. (B) Quantitative analysis of KIF13A- FRB-Myc band intensity after immunoprecipitation with Hrs or STAM, expressed as intensity normalized to mCh control. Hrs precipitates KIF13A more after high K+ treatment. Results are preliminary and represent only 1 week of replicates.

Knockdown of KIF13A ablates activity-dependent increase in ESCRT-0 motility and slows degradation of SV proteins

To test whether KIF13A is necessary for the transport of ESCRT-0 vesicles in neurons, we next examined the effects of KIF13A knockdown on ESCRT-0 protein axonal transport and function. For these studies, we used an shRNA against rat KIF13A (sh13A) that produced a

~60% knockdown of the protein after lentiviral transduction (transduced 7 DIV, harvested 14

DIV; Figure 20A-B). Our previous work showed that a direct Hrs knockdown inhibits the

degradation of several SV membrane proteins, including Synaptic Vesicle protein 2 (SV2) and

Vesicle-associated membrane protein 2 (VAMP2)(Sheehan et al. 2016). We therefore reasoned that knockdown of KIF13A should similarly impair the degradation of these SV proteins if it is required for anterograde axonal transport of Hrs. To test this, we used our previously described cycloheximide-chase assay (Sheehan et al. 2016) to measure the degradation of SV2 and

VAMP2 in neurons expressing sh13A or a scrambled control shRNA (shCtrl). We found that degradation of both SV2 and VAMP2 was reduced in the presence of sh13A compared to shCtrl

(Figure 21A-C), suggesting that KIF13A is necessary to the timely breakdown of these SV proteins. To directly examine the effects of KIF13A knockdown on the axonal motility of Hrs, we co-expressed EGFP-Hrs together with sh13A or shCtrl in neurons cultured in microfluidic chambers as previously described. Time-lapse imaging revealed that sh13A expression did not alter either the motility (Figure 20C), directional movement (Figure 20C), or speed (Figure 20D) of Hrs under control conditions compared to shCtrl. However, this hairpin completely abolished the activity-dependent increase in Hrs motility (Figure 22A,C,E). Similar effects were seen for

STAM1, as baseline puncta motility was unaffected by sh13A, while activity-dependent motility was inhibited (Figure 22B,D,F). Overall, these data indicate that KIF13A is responsible for the activity-dependent transport of ESCRT-0+ vesicles in axons and is necessary for the degradation of SV protein ESCRT cargo.

Figure 20. Knockdown of KIF13A by sh13A expression in neurons. (A) Representative immunoblot of lysates of 13-15 DIV hippocampal neurons transduced with sh13A or scrambled shRNA control (shCtrl) on 7 DIV, probed with antibodies against KIF13A and tubulin. (B) Quantification of KIF13A band intensity, normalized to tubulin control, with shCtrl or sh13A expression (**P<0.01, unpaired t-test; ≥3 separate experiments). Bars show mean ± SEM. (C) Breakdown of directional movement of Hrs puncta under control conditions, expressed as percentage of total Hrs puncta, with shCtrl or sh13A expression. Knockdown of KIF13A does not alter baseline Hrs directional movement. (D) Average speed of motile Hrs puncta (µm/sec) in the presence of shCtrl or sh13A.

Summary

These results strongly implicate the kinesin-3 motor KIF13A in the activity-dependent anterograde motility of ESCRT-0. The finding that knockdown of KIF13A slows SV protein degradation (Figure 21A-C) also suggests that this activity-dependent motility may be at play, undetectably to our methods, even in the lower-activity baseline condition of our cultured

neurons. This and many other unknowns, including the identities of the motors and adaptors involved in non-activity-dependent ESCRT-0 motility, comprise significant questions to be addressed in future work. However, these data reveal a vital link between the activity-dependent increase in motility that we find is specific to ESCRT-0, and the degradation of SV proteins, the manifest function of the ESCRT pathway in neurons. These results furthermore ideate KIF13A as a vital actor in the maintenance of synaptic proteostasis, and therefore a potentially crucial pillar of long-term neuronal health. However, a more complete picture of the mechanism by which activity regulates the interaction between KIF13A and ESCRT-0 will be necessary to fully substantiate these findings. Overall, this work represents the first identification of a kinesin motor protein essential to the activity-dependent motility of the newly described specialized

ESCRT-0+/Rab5+ endosomes. Future work will continue to characterize the function of these structures, their interaction with synapses, and the regulation of their motility – activity- dependent and otherwise.

Figure 21. Knockdown of KIF13A slows SV protein degradation. (A) Representative immunoblots of SV2 and VAMP2, with corresponding tubulin loading controls, from lysates of 14 DIV hippocampal neurons transduced with shCtrl or sh13A to knockdown KIF13A and treated for 24 h with DMSO control (CON) or cycloheximide (CHX). (B-C) Quantification of the fold-change in SV2 (C) and VAMP2 (D) degradation under these conditions, calculated as previously described (Sheehan et al. 2016). Degradation of both proteins is significantly decreased in the presence of sh13A (*P<0.05, unpaired t-test; ≥3 separate experiments, n = 5 (SV2), 11 (VAMP2) cell lysates/condition).

Figure 22. Knockdown of KIF13A inhibits activity-dependent transport of ESCRT-0 proteins. (A) Representative images and corresponding kymographs of microfluidically isolated hippocampal axons transduced with EGFP-Hrs and shCtrl under control conditions (upper panels) and after 2 h treatment with Bic/4AP (lower panels). Scale bar, 10 µm. (B) Representative images and

corresponding kymographs of microfluidically isolated hippocampal axons expressing STAM1-mCh and shCtrl under control conditions (upper panels) and after 2 h treatment with Bic/4AP (lower panels). Scale bar, 10 µm. (C) Representative images and corresponding kymographs of microfluidically isolated hippocampal axons expressing EGFP-Hrs and sh13A under control conditions (upper panels) and after 2 h treatment with Bic/4AP (lower panels). Scale bar, 10 µm. (D) Representative images and corresponding kymographs of microfluidically isolated hippocampal axons expressing STAM1-mCh and sh13A under control conditions (upper panels) and after 2 h treatment with Bic/4AP (lower panels). Scale bar, 10 µm. (E) Percentage of motile Hrs puncta in axons, showing that the increased motility of Hrs after 2 h Bic/4AP treatment is abolished in sh13A-expressing neurons (***P<0.001, one-way ANOVA with Sidak’s multiple comparisons test; ≥3 separate experiments, n=14 (shCtrl+control), 16 (shCtrl+2 h Bic/4AP), 16 (sh13A+control), 15 (sh13A+2 h Bic/4AP) videos/condition). (F) Percentage of motile STAM1 puncta in axons, showing that the increased motility of STAM1 after 2 h Bic/4AP treatment is abolished in sh13A-expressing neurons (***P<0.001, one-way ANOVA with Sidak’s multiple comparisons test; ≥3 separate experiments, n=14 (shCtrl+control), 15 (shCtrl+2 h Bic/4AP), 13 (sh13A+control), 13 (sh13A+2 h Bic/4AP) videos/condition). All scatter plots show mean ± SEM.

Chapter 7: Discussion and Future Directions

In this study, we demonstrate a novel mechanism for spatiotemporal regulation of the

ESCRT pathway in neurons. Our findings show that axonal vesicles carrying ESCRT-0 proteins exhibit increased anterograde and bidirectional motility in response to neuronal firing (Figures

7&23), providing the first example of activity-dependent transport of degradative machinery in axons. Previous investigations have reported that both the axonal motility of mitochondria and the inter-bouton transport of SVs are regulated by neuronal activity in hippocampal neurons, specifically by presynaptic calcium influx (M.-Y. Lin and Sheng 2015; Chen and Sheng 2013;

Gramlich and Klyachko 2017; Qu et al. 2019). These processes are hypothesized to contribute to the activity-dependent remodeling of synapses in order to alter neurotransmitter release properties, and to ensure that highly active synapses are properly replenished (Gramlich &

Klyachko, 2017; Lin & Sheng, 2015). Synaptic activity has also been shown to induce the postsynaptic recruitment of proteasomes and lysosomes in cultured neurons, as well as the local biogenesis of autophagosomes at pre- and postsynaptic sites (T. Wang et al. 2015; Bingol and

Schuman 2006; Goo et al. 2017; Shehata et al. 2012; Soukup et al. 2016; Vanhauwaert et al.

2017; Kulkarni and Maday 2018a). However, the axonal transport of autophagosomes in cultured neurons appears to be a constitutive process that is unaffected by stimuli such as synaptic activity and glucose levels (Maday and Holzbaur 2016; Maday, Wallace, and Holzbaur 2012). Axonal autophagosomes, once formed at the presynapse, undergo reliable retrograde trafficking to the soma while acidifying into autolysosomes (Maday and Holzbaur 2016; Maday, Wallace, and

Holzbaur 2012). In contrast, we find that axonal transport of ESCRT-0 proteins Hrs and STAM1 is bidirectional and sensitive to neuronal activity levels. We hypothesize that these characteristics facilitate increased contact of this early ESCRT machinery with SV pools, thereby catalyzing the

activity-dependent degradation of SV proteins. This regulation by neuronal firing is undoubtedly mediated by a consequence of an action potential traveling along an axon and invading a bouton, i.e. calcium influx, rather than by neurotransmission itself. Calcium chelators could be used in the presence of bicuculline/4AP to test whether calcium mediates the activity-dependent changes to ESCRT motility, as is the case for mitochondrial and SV motility. One could also test whether pharmacologically inducing an influx of extracellular calcium or an efflux of stored calcium from the endoplasmic reticulum induces ESCRT motility. If this motility is calcium-dependent, it should next be determined whether calcium stimulates, or is necessary for, the interaction between ESCRT-0 and KIF13A, as described in detail below.

Figure 23. Model of how activity stimulates the anterograde transport of ESCRT-0+ vesicles. We hypothesize that an increase in intra-axonal Ca2+ due to neuronal firing stimulates the association of axonal ESCRT-0-positive vesicles with plus-end directed kinesin KIF13A. Over time, this leads to an

increase in anterograde transport of these vesicles, which deliver Hrs and STAM1 to SV pools in order to mediate the activity-dependent degradation of SV proteins.

Autophagy and the ESCRT pathway

My work shows that the axonal motility patterns of specific ESCRT-0 and -I proteins differ significantly from that of autophagosome marker LC3. However, the complex and interdependent relationship between autophagy and the ESCRT pathway precludes the conclusion that these pathways are therefore uncoupled in the axon. It is clear from the literature that autophagy is dysregulated in the absence of functional ESCRT, particularly ECRT-III proteins (J.-A. Lee et al. 2007; Tor Erik Rusten and Simonsen 2008). These observations have kindled several theories about the possible function of the ESCRT pathway in autophagy induction, closure of the phagophore, and/or fusion with the late endosome/lysosome (T E

Rusten and Stenmark 2009). In axons, Cheng et. al. found that autophagosomes must fuse with late endosomes in order to acquire enough dynein to begin their retrograde journey to the soma

(Cheng et al. 2015). Moreover, it has been shown repeatedly that axonal autophagosomes form in growth cones or at presynaptic boutons (Maday, Wallace, and Holzbaur 2012; T. Wang et al.

2015; Maday et al. 2014), obviating their need to travel in the anterograde direction to interact with synaptic cargoes. These findings, in concert with our conclusion that ESCRT membrane assemblies are likely formed in the soma, suggest that ESCRT proteins may show a more complex directional profile than autophagosome markers simply because they must travel both to and from the synapse. Moreover, given the frequent and potentially necessary fusion of late endosomes with autophagosomes, there is a strong possibility that retrogradely-moving axonal

ESCRT+ vesicles are amphisomes or precursors to autolysosomes. To test this, neurons co- transfected with ESCRT proteins and LC3 could be subjected to live axonal imaging, and the

directional breakdown of these co-positive structures compared to those of individual ESCRT proteins and LC3 alone. It is possible that ESCRT proteins associated with late endosomes, particularly ESCRT-III, will show a stronger colocalization with autophagosome markers and more retrograde motility. Moreover, additional correlative light electron microscopy (CLEM) experiments with ESCRT-I, -II, and -III proteins should be performed to determine whether any

ESCRT+ structures exhibit morphological hallmarks of amphisomes. These experiments will elucidate the relationship between ESCRTs and autophagosomes in the axon.

Specificity of Activity-Dependent Regulation to ESCRT-0+ Vesicles

Although there are multiple ESCRT components, their roles are highly interconnected, as illustrated by the fact that mutation of a single ESCRT protein is enough to thoroughly damage endolysosomal processing and induce neurodegeneration (D. Lee et al. 2019; J. A. Lee et al.

2007; Y. Zhang et al. 2017; Muzioł et al. 2006). It is evident from the literature that ESCRT function relies on the coordinated recruitment of all the ESCRT complexes to the same microdomain on the endosomal membrane (Henne, Buchkovich, and Emr 2011; Hurley 2015).

Moreover, our previous study indicated that both ESCRT-0 component Hrs and ESCRT-III component CHMP2B are recruited to presynaptic sites in response to prolonged neuronal firing

(24 h Bic/4AP; (Sheehan et al. 2016)). In light of these findings, it is interesting that the regulation of ESCRT-0 transport by neuronal activity does not appear to extend to the ESCRT-I complex (Figures 7&8). However, an analysis of the unique role played by ESCRT-0 in overall

ESCRT function suggests why this may be the case. First, of all the ESCRT complexes, ESCRT-

0 alone is capable of stable, independent membrane association. This feature makes it easier for us to characterize the axonal motility of ESCRT-0 proteins in our overexpression system and

provides a potential explanation as to why the regulation of ESCRT-0 axonal transport by activity appears more robust than for downstream ESCRT complexes. It is possible that regulation of axonal transport vesicles carrying ESCRT-0 is sufficient for localization of the entire pathway, if ESCRT-0 can initiate recruitment of the rest of the pathway. This concept underlies the second reason why the ESCRT-0 complex alone may be regulated by activity.

While downstream ESCRT complexes interact mainly with each other and fleetingly with the membrane, ESCRT-0 shows the strongest interaction with ubiquitin, and thus with nascent

ESCRT-0 cargo. This property is consistent with our finding that Hrs interacts with SV proteins for longer time periods in the presence of a DUB inhibitor to boost ubiquitination (Figure

14B,D), and suggests that the ESCRT-0 complex alone has the ability to ‘discover’ and interact with cargoes. Altogether, this leads us to hypothesize that ESCRT-0 vesicles are the ‘first responders’ of the ESCRT pathway, and therefore the most sensitive to regulatory stimuli. In this case, the targeting of ESCRT-0 proteins to presynaptic sites, perhaps via an interaction with ubiquitinated synaptic vesicle cargo, is sufficient to localize and activate the rest of the ESCRT pathway. This concept must be tested by characterizing the motility and synaptic localization of downstream ESCRT pathway components in the presence or absence of Hrs and activity. The

‘first-responder’ theory will be bolstered if Hrs knockdown impedes both ESCRT pathway function (as previously shown; (Sheehan et al. 2016)) and the synaptic localization of later

ESCRT proteins. If our hypothesis is borne out by these experiments, the next question to address is how downstream ESCRT complexes are localized to the sites ‘chosen’ by ESCRT-0.

Are vesicles carrying ESCRT-I, -II, and -III proteins directed to presynaptic sites by activity, but in a time course outside the scope of this study (2h of stimulation)? Or are such vesicles present diffusely throughout axons, and thus already available for recruitment wherever ESCRT-0

complexes go? These questions can be answered by mapping the localization and transport properties of ESCRT-I, -II, and -III components at different time points after treatment with

Bic/4AP. These data will be vital to expanding our understanding of ESCRT pathway function and regulation in neurons.

Identity of ESCRT-0+ Vesicles

Our data indicate that axonal ESCRT-0+ vesicles represent a specialized endosomal subtype. Endosomes are highly dynamic and heterogenous structures in all cells, and particularly in neurons, whose complex morphology and high protein trafficking demands have given rise to multiple distinct classes of endosomes that are spatially restricted within different neuronal compartments (Maday and Holzbaur 2016; Yap et al. 2018). Early endosomes are classically defined by the markers Rab5 and EEA1 (Langemeyer, Fröhlich, and Ungermann 2018;

Simonsen et al. 1998; Gorvel et al. 1991); however, recent studies have identified sub-categories of early endosomes in neurons that can be differentiated both by their markers and axonal transport patterns (Olenick, Dominguez, and Holzbaur 2018; Leonard et al. 2008). The axonal

ESCRT-0+ vesicles we describe represent a novel addition to this category. We find that while the total pool of axonal Rab5+ vesicles exhibits high motility under baseline conditions and a relatively modest activity-induced increase in anterograde motility (Figure 10F), the Hrs+ subset of this pool exhibits limited motility at baseline, but significant activity-induced anterograde and bidirectional motility (Figure 7F). Moreover, our CLEM experiments demonstrate that axonal

Hrs+ vesicles can be degradative structures (Figure 15A), although we cannot yet correlate their morphology with their axonal transport history to determine what stage of the degradative process they represent. In future experiments, we will track activity-regulated Hrs+ vesicles over a longer time course to assess: 1) where these vesicles originate (i.e. in the soma or proximal

axon), 2) which motor proteins transport these vesicles, and how neuronal activity affects their vesicle association, and 3) how vesicle morphology correlates with transport history (e.g. do motile structures have a different morphology from stationary structures). This will require imaging axonal ESCRT-0+ vesicles over a longer period of time, as well as marking vesicles in fluidically isolated neurons in either the somatodendritic or axonal compartments, and determining their localization and morphology at different time points. These results will provide crucial insights into the regulation and identity of the activity-responsive Hrs vesicles characterized here. Moreover, STAM1 is also is a member of the ESCRT-0 complex. We did not assess STAM1 vesicle morphology or colocalization with Rab5 in this study, but we have previously observed that co-expressed Hrs and STAM1 are highly colocalized in axons, as the literature predicts (Mayers et al. 2011; Ren and Hurley 2010), and we find here that STAM1+ vesicles exhibit very similar activity-dependent dynamics to Hrs+ vesicles (Figure 7). This unique sensitivity of ESCRT-0+ vesicles to neuronal activity, together with their morphological characteristics and ubiquitin-dependent colocalization with SV pools (Figure 14B,D), suggests that they have a specific function: mediating the activity-dependent degradation of old and/or dysfunctional SV proteins. This hypothesis aligns with previous research indicating that SV proteins become increasingly dysfunctional as they age and undergo multiple cycles of exo/endocytosis (Truckenbrodt et al. 2018; Fernandes et al. 2014). This theory can be further tested by using fixed and live-cell imaging to assess the relative colocalization of ESCRT-0 with

SV proteins of different ages (as determined in (Truckenbrodt et al. 2018)), as well as whether the pausing of ESCRT-0 at presynapses is affected by the proportion of older vesicles. We hypothesize that older synaptic vesicle proteins, which have had longer to accumulate molecular damage, are more likely to be ubiquitinated, and thus more likely to attract ESCRT-0. Our work

lays out the possibility that the neuronal ESCRT pathway is designed to deal with this issue of aged and dysfunctional SV proteins and is activated by high neuronal firing in order to protect synaptic function.

Hrs Endosomal Targeting

In addition to characterizing the dynamic behavior and morphology of ESCRT-0 axonal transport vesicles, this study investigates the mechanisms of Hrs targeting to these vesicles. As in non-neuronal cells (Gaullier et al., 2000; Komada & Soriano, 1999), we find that the association of Hrs with vesicles is dependent upon interactions between its FYVE domain and the endosomal lipid PI3P (Figure 13). Blocking PI3P synthesis with a VPS34 inhibitor (SAR405) or expressing a FYVE domain-inactivating mutant form of Hrs, leads to a significant loss of Hrs membrane targeting in the somatodendritic compartment (Figure 13A-D). However, treatment with SAR405 does not alter the number, speed, or motility of Hrs puncta in axons (Figure 13E-

I). Together, these findings indicate that the axonal pool of Hrs-positive vesicles is assembled in the somatodendritic compartment, and that Hrs interactions with axonal transport vesicles are stable for at least 2 hours. A potential alternative explanation for this finding is that a different domain of Hrs mediates its association with axonal vesicles. Indeed, it was previously reported that the FYVE domain is insufficient for Hrs membrane targeting, as the C-terminal coiled-coil domain is also necessary for this function (Raiborg et al. 2001). Interestingly, the coiled-coil domain of Hrs is known to interact with SNAP-25 (Kwong et al. 2000), a SV-associated protein and component of the SNARE complex required for SV fusion (McMahon & Südhof, 1995;

Söllner et al., 1993). Given that SNAP-25 must also be transported to presynaptic terminals, it is conceivable that its presence on axonal transport vesicles is required for the recruitment and/or

stabilization of Hrs on these structures. This can be assessed by determining via imaging whether

SNAP-25 is present on ESCRT-0+ vesicles, and if so whether its knockdown or mutation alters the stability of Hrs on endosomal membranes. It is also interesting to note that EEA1, which targets to early endosome membranes via its FYVE domain, does not localize to axons (Gaullier et al., 2000; Lawe et al., 2000; Wilson et al., 2000), indicating that additional domains and protein interactions may be required for axonal targeting of FYVE domain-containing proteins.

These domains could be identified by selectively mutating Hrs and assessing membrane localization of the mutants in both the somatodendritic and axonal compartments. This work will continue to define the interactions that direct axonal Hrs to vesicular membranes, and such findings will likely inform our understanding of the axonal targeting of other proteins.

Interestingly, recent studies have uncovered the prominent role that the axon initial segment

(AIS) plays in maintaining separation between axonal and somatodendritic pools of vesicles

(Kulkarni and Maday 2018a; Song et al. 2009). It would be fascinating to address whether this physical barrier plays a role in regulating the axonal transport of ESCRT proteins, and whether it is influenced by activity. This could be done by assessing ESCRT subcellular localization in the presence of a knockdown or mutation of a crucial AIS protein such as Ankrin-G, and by determining via imaging whether there are any changes to how ESCRT proteins interact with the

AIS in the presence of high neuronal activity.

Activity-Dependent Transport of ESCRT-0+ Vesicles

Finally, our work has begun to elucidate the mechanisms of activity-dependent transport of ESCRT-0+ vesicles, in particular the role of kinesin-3 family member KIF13A in this process.

Surprisingly, we find that although both KIF13A and KIF13B mediate transport of Rab5 in a

reconstituted transport assay (Figure 17C-F), as previously reported (Bentley et al., 2015), only

KIF13A transports Hrs (Figure 18A-D). This specificity is confirmed by coimmunoprecipitation assays in which Hrs precipitates KIF13A to a much greater degree than KIF13B (Figure 18E-F), further bolstering the theory that ESCRT-0 vesicles comprise a specialized subset of pro- degradative endosomes with unique transport properties. KIF13A and KIF13B are highly homologous proteins that share multiple cargoes, including early and recycling endosome associated proteins Rab5, Rab10, and transferrin receptor (Jenkins et al., 2012; Bentley et al.,

2015; Etoh & Fukuda, 2019). These kinesins also have distinct cargoes, including low density lipoprotein receptor (for KIF13B; Jenkins et al., 2012) and serotonin type 1a receptor (for

KIF13A; Zhou et al., 2013), as well as Hrs and STAM1 (Figure 18E-F). It is possible that these distinct cargoes reflect differences in the subcellular localization of KIF13A and KIF13B in neurons. However, current reports about their localization are conflicting, with one study showing that KIF13A localizes to axonal and dendritic vesicles while KIF13B is primarily dendritic (Jenkins et al., 2012), and another showing that KIF13B has an axonal bias whereas

KIF13A is exclusively dendritic (Yang et al., 2019). In our experiments, we find that knockdown of KIF13A prevents the activity-induced motility of Hrs and STAM1 in axons (Figure 22), indicating that KIF13A has an essential role in the axonal transport of these ESCRT-0 proteins.

Our findings also highlight the importance of KIF13A-mediated transport in SV protein degradation, which is impaired by KIF13A knockdown under basal conditions (Figure 21).

These data suggest that activity-dependent transport of ESCRT-0 proteins occurs even under conditions of lower neuronal firing but may be enveloped into what we identified as ‘baseline’ motility. As delineated above, additional experiments are needed to characterize the mechanism by which KIF13A facilitates the activity-dependent transport of Hrs, we well as its interaction

with ESCRT-0 vesicles. The first experiment that should be conducted is live cell imaging of neurons co-transfected with Hrs and KIF13A to determine what percentage of axonal Hrs+ vesicles interact with KIF13A, how the presence or absence of KIF13A on an Hrs+ vesicle correlates to its motility pattern, and whether KIF13A is present on only activity-recruited Hrs+ vesicles. It is possible that activity increases the colocalization between Hrs and KIF13A, which can be tested by proximity ligation assay in neurons co-transfected with Hrs and KIF13A and treated with Bic/4AP. This would also reveal the sub-cellular location of Hrs when the interaction occurs. If activity does increase Hrs/KIF13A interaction, it would be interesting to assess how KIF13A is localized to Hrs+ vesicles, whether by fusion between these vesicles and others carrying KIF13A, or by a calcium-dependent recruitment of cytoplasmic KIF13A to these vesicles. To distinguish between these possibilities, axons of neurons co-transfected with

KIF13A and Hrs should be imaged in the presence and absence of activity treatments to assess the localization of KIF13A relative to Hrs+ vesicles. It is also possible that KIF13A is always present on Hrs+ vesicles, but is activated, or recruits a necessary effector, in response to activity.

This scenario would be supported if the described experiments do not show an increase in

Hrs/KIF13A colocalization in response to neuronal activity. Mass spectrometry and proteomic analysis could be performed on purified Hrs+ and KIF13A+ vesicles to determine whether their composition changes with activity. Future experiments will also identify other motor proteins responsible for ESCRT-0 vesicle motility, including the dynein adaptors involved in retrograde transport of these vesicles. These can be initially identified using immunoprecipitation and split- kinesin assays as described here. These studies are particularly important in light of the bidirectional transport profile of ESCRT-0 proteins (Figure 7F,M). As mentioned above, it is possible that the retrogradely-motile ESCRT-0 puncta we observe represent later stage

endosomes or amphisomes that have obtained their cargo and begun the journey back to the soma for lysosomal degradation. If this is the case, we may see that later ESCRT proteins (-I, -II,

-III, VPS4) are also present on these retrogradely-moving structures. Overall, there is still much to learn about the roles and regulation of the ESCRT pathway in neurons. However, this work provides the first characterization of the spatial regulation of ESCRT components in axons.

Conclusion

The ESCRT pathway plays a vital role in maintaining neuronal health and is intimately linked to neurodegenerative disease. Here, we have illuminated a novel mechanism for spatiotemporal regulation of the ESCRT pathway in neurons, based on the activity-induced axonal transport of pro-degradative ESCRT-0 vesicles by kinesin motor protein KIF13A. These findings broaden our understanding of how protein degradation is regulated in neurons and shed light on how disruption of this degradative pathway may contribute to the etiology of neurodegenerative disease.

References

Ashrafi, Ghazaleh, Julia S. Schlehe, Matthew J. LaVoie, and Thomas L. Schwarz. 2014. “Mitophagy of Damaged Mitochondria Occurs Locally in Distal Neuronal Axons and Requires PINK1 and Parkin.” Journal of Cell Biology. https://doi.org/10.1083/jcb.201401070.

Babst, Markus, David J. Katzmann, Eden J. Estepa-Sabal, Timo Meerloo, and Scott D. Emr. 2002. “ESCRT-III: An Endosome-Associated Heterooligomeric Protein Complex Required for MVB Sorting.” Developmental Cell. https://doi.org/10.1016/S1534-5807(02)00220-4.

Babst, Markus, Beverly Wendland, Eden J. Estepa, and Scott D. Emr. 1998. “The Vps4p AAA ATPase Regulates Membrane Association of a Vps Protein Complex Required for Normal Endosome Function.” EMBO Journal. https://doi.org/10.1093/emboj/17.11.2982.

Bache, Kristi G., Andreas Brech, Anja Mehlum, and Harald Stenmark. 2003. “Hrs Regulates Multivesicular Body Formation via ESCRT Recruitment to Endosomes.” Journal of Cell Biology. https://doi.org/10.1083/jcb.200302131.

Bache, Kristi G, Camilla Raiborg, Anja Mehlum, and Harald Stenmark. 2003. “STAM and Hrs Are Subunits of a Multivalent Ubiquitin-Binding Complex on Early Endosomes.” Journal of Biological Chemistry 278 (14): 12513–21. https://doi.org/10.1074/jbc.M210843200 .

Bajorek, Monika, Heidi L. Schubert, John McCullough, Charles Langelier, Debra M. Eckert, William May B. Stubblefield, Nathan T. Uter, David G. Myszka, Christopher P. Hill, and Wesley I. Sundquist. 2009. “Structural Basis for ESCRT-III Protein Autoinhibition.” Nature Structural and Molecular Biology. https://doi.org/10.1038/nsmb.1621.

Bentley, Marvin, and Gary Banker. 2015. “A Novel Assay to Identify the Trafficking Proteins That Bind to Specific Vesicle Populations.” Current Protocols in Cell Biology 69 (1): 13.8.1-13.8.12. https://doi.org/10.1002/0471143030.cb1308s69.

Bentley, Marvin, Helena Decker, Julie Luisi, and Gary Banker. 2015. “A Novel Assay Reveals Preferential Binding between Rabs, Kinesins, and Specific Endosomal Subpopulations.” Journal of Cell Biology 208 (3): 273–81. https://doi.org/10.1083/jcb.201408056.

Berg, Trond Olav, Monica Fengsrud, Per Eivind Strømhaug, Trond Berg, and Per O Seglen. 1998. “Isolation and Characterization of Rat Liver Amphisomes: EVIDENCE FOR FUSION OF AUTOPHAGOSOMES WITH BOTH EARLY AND LATE ENDOSOMES.” Journal of Biological Chemistry 273 (34): 21883–92. https://doi.org/10.1074/jbc.273.34.21883.

Bezprozvanny, Ilya, and Peter Robin Hiesinger. 2013. “The Synaptic Maintenance Problem: Membrane Recycling, Ca2+homeostasis and Late Onset Degeneration.” Molecular Neurodegeneration 8 (1): 23. https://doi.org/10.1186/1750-1326-8-23.

Bilodeau, Patricia S., Stanley C. Winistorfer, William R. Kearney, Andrew D. Robertson, and Robert C. Piper. 2003. “Vps27-Hse1 and ESCRT-I Complexes Cooperate to Increase Efficiency of Sorting Ubiquitinated Proteins at the Endosome.” Journal of Cell Biology. https://doi.org/10.1083/jcb.200305007.

Bingol, Baris, and Erin M Schuman. 2006. “Activity-Dependent Dynamics and Sequestration of Proteasomes in Dendritic Spines.” Nature 441 (7097): 1144–48. https://doi.org/10.1038/nature04769.

Birdsall, Veronica, Jose C Martinez, Lisa Randolph, Ulrich Hengst, and Clarissa L Waites. 2019. Live Imaging of ESCRT Proteins in Microfluidically Isolated Hippocampal Axons BT - The ESCRT Complexes: Methods and Protocols. Edited by Emmanuel Culetto and Renaud Legouis. New York, NY: Springer New York. https://doi.org/10.1007/978-1-4939-9492- 2_9.

Birdsall, Veronica, and Clarissa L Waites. 2019. “Autophagy at the Synapse.” Neuroscience Letters 697: 24–28. https://doi.org/https://doi.org/10.1016/j.neulet.2018.05.033.

Borland, Helena, and Frederik Vilhardt. 2017. “Prelysosomal Compartments in the Unconventional Secretion of Amyloidogenic Seeds.” International Journal of Molecular Sciences . https://doi.org/10.3390/ijms18010227.

Bowers, Katherine, Siân C. Piper, Melissa A. Edeling, Sally R. Gray, David J. Owen, Paul J. Lehner, and J. Paul Luzio. 2006. “Degradation of Endocytosed Epidermal Growth Factor and Virally Ubiquitinated Major Histocompatibility Complex Class I Is Independent of Mammalian ESCRTII.” Journal of Biological Chemistry. https://doi.org/10.1074/jbc.M508632200.

Chen, Yanmin, and Zu Hang Sheng. 2013. “Kinesin-1-Syntaphilin Coupling Mediates Activity- Dependent Regulation of Axonal Mitochondrial Transport.” Journal of Cell Biology. https://doi.org/10.1083/jcb.201302040.

Cheng, Xiu-Tang, Yu-Xiang Xie, Bing Zhou, Ning Huang, Tamar Farfel-Becker, and Zu-Hang Sheng. 2018. “Characterization of LAMP1-Labeled Nondegradative Lysosomal and Endocytic Compartments in Neurons.” Journal of Cell Biology 217 (9): 3127–39. https://doi.org/10.1083/jcb.201711083.

Cheng, Xiu-Tang, Bing Zhou, Mei-Yao Lin, Qian Cai, and Zu-Hang Sheng. 2015. “Axonal Autophagosomes Recruit Dynein for Retrograde Transport through Fusion with Late Endosomes.” Journal of Cell Biology 209 (3): 377–86. https://doi.org/10.1083/jcb.201412046.

Chi, Hao, Hui-Yun Chang, and Tzu-Kang Sang. 2018. “Neuronal Cell Death Mechanisms in Major Neurodegenerative Diseases.” International Journal of Molecular Sciences 19 (10): 3082. https://doi.org/10.3390/ijms19103082.

Christoforidis, Savvas, Marta Miaczynska, Keith Ashman, Matthias Wilm, Liyun Zhao, Shu- Chin Yip, Michael D Waterfield, Jonathan M Backer, and Marino Zerial. 1999. “Phosphatidylinositol-3-OH Kinases Are Rab5 Effectors.” Nature Cell Biology 1 (4): 249– 52. https://doi.org/10.1038/12075.

Cohen, Laurie D, and Noam E Ziv. 2017. “Recent Insights on Principles of Synaptic Protein Degradation.” F1000Research 6 (May): 675. https://doi.org/10.12688/f1000research.10599.1.

Deinhardt, Katrin, Sara Salinas, Carole Verastegui, Rose Watson, Daniel Worth, Sarah Hanrahan, Cecilia Bucci, and Giampietro Schiavo. 2006. “Rab5 and Rab7 Control Endocytic Sorting along the Axonal Retrograde Transport Pathway.” Neuron 52 (2): 293– 305. https://doi.org/10.1016/j.neuron.2006.08.018.

Ding, Wen Xing, and Xiao Ming Yin. 2012. “Mitophagy: Mechanisms, Pathophysiological Roles, and Analysis.” Biological Chemistry. https://doi.org/10.1515/hsz-2012-0119.

Ebrahimi-Fakhari, Darius, Afshin Saffari, Lara Wahlster, Alessia Di Nardo, Daria Turner, Tommy L. Lewis, Christopher Conrad, et al. 2016. “Impaired Mitochondrial Dynamics and Mitophagy in Neuronal Models of Tuberous Sclerosis Complex.” Cell Reports. https://doi.org/10.1016/j.celrep.2016.09.054.

Ehlers, Michael D. 2000. “Reinsertion or Degradation of AMPA Receptors Determined by Activity-Dependent Endocytic Sorting.” Neuron. https://doi.org/10.1016/S0896- 6273(00)00129-X.

Esposito, Giovanni, Fernandes Ana Clara, and Patrik Verstreken. 2012. “Synaptic Vesicle Trafficking and Parkinson’s Disease.” Developmental Neurobiology 72 (1): 134–44. https://doi.org/10.1002/dneu.20916.

Etoh, Kan, and Mitsunori Fukuda. 2019. “Rab10 Regulates Tubular Endosome Formation through KIF13A and KIF13B Motors.” Journal of Cell Science 132 (5): jcs226977. https://doi.org/10.1242/jcs.226977.

Fader, C M, and M I Colombo. 2008. “Autophagy and Multivesicular Bodies: Two Closely Related Partners.” Cell Death And Differentiation 16: 70. https://doi.org/10.1038/cdd.2008.168.

Feng, Qiong, Yu Luo, Xiang-Nan Zhang, Xi-Fei Yang, Xiao-Yue Hong, Dong-Shen Sun, Xia- Chun Li, et al. 2020. “MAPT/Tau Accumulation Represses Autophagy Flux by Disrupting IST1-Regulated ESCRT-III Complex Formation: A Vicious Cycle in Alzheimer Neurodegeneration.” Autophagy 16 (4): 641–58. https://doi.org/10.1080/15548627.2019.1633862.

Ferguson, Shawn M. 2018. “Axonal Transport and Maturation of Lysosomes.” Current Opinion in Neurobiology 51 (August): 45–51. https://doi.org/10.1016/j.conb.2018.02.020.

Fernandes, Ana Clara, Valerie Uytterhoeven, Sabine Kuenen, Yu-Chun Wang, Jan R Slabbaert, Jef Swerts, Jaroslaw Kasprowicz, Stein Aerts, and Patrik Verstreken. 2014. “Reduced Synaptic Vesicle Protein Degradation at Lysosomes Curbs TBC1D24/Sky-Induced Neurodegeneration.” Journal of Cell Biology 207 (4): 453–62. https://doi.org/10.1083/jcb.201406026.

Ferreira, Joana S., Jeannette Schmidt, Pedro Rio, Rodolfo Águas, Amanda Rooyakkers, Ka Wan Li, August B. Smit, Ann Marie Craig, and Ana Luisa Carvalho. 2015. “GluN2B-Containing NMDA Receptors Regulate AMPA Receptor Traffic through Anchoring of the Synaptic Proteasome.” Journal of Neuroscience. https://doi.org/10.1523/JNEUROSCI.3567-14.2015.

Fischer von Mollard, G., B. Stahl, C. Walch-Solimena, K. Takei, L. Daniels, A. Khoklatchev, P. De Camilli, T. C. Sudhof, and R. Jahn. 1994. “Localization of Rab5 to Synaptic Vesicles Identifies Endosomal Intermediate in Synaptic Vesicle Recycling Pathway.” European Journal of Cell Biology.

Flores-Rodriguez, Neftali, David A Kenwright, Pei-Hua Chung, Andrew W Harrison, Flavia Stefani, Thomas A Waigh, Victoria J Allan, and Philip G Woodman. 2015. “ESCRT-0 Marks an APPL1-Independent Transit Route for EGFR between the Cell Surface and the EEA1-Positive Early Endosome.” Journal of Cell Science 128 (4): 755 LP – 767. https://doi.org/10.1242/jcs.161786.

Gaullier, Jean-Michel, Eva Rønning, David J Gillooly, and Harald Stenmark. 2000. “Interaction of the EEA1 FYVE Finger with Phosphatidylinositol 3-Phosphate and Early Endosomes: ROLE OF CONSERVED RESIDUES .” Journal of Biological Chemistry 275 (32): 24595– 600. https://doi.org/10.1074/jbc.M906554199.

Goo, Marisa S., Samantha L. Scudder, and Gentry N. Patrick. 2015. “Ubiquitin-Dependent Trafficking and Turnover of Ionotropic Glutamate Receptors.” Frontiers in Molecular Neuroscience. https://doi.org/10.3389/fnmol.2015.00060.

Goo, Marisa S, Laura Sancho, Natalia Slepak, Daniela Boassa, Thomas J Deerinck, Mark H Ellisman, Brenda L Bloodgood, and Gentry N Patrick. 2017. “Activity-Dependent Trafficking of Lysosomes in Dendrites and Dendritic Spines.” Journal of Cell Biology 216 (8): 2499–2513. https://doi.org/10.1083/jcb.201704068.

Gorman, Adrienne M. 2008. “Neuronal Cell Death in Neurodegenerative Diseases: Recurring Themes around Protein Handling.” Journal of Cellular and Molecular Medicine 12 (6A): 2263–80. https://doi.org/10.1111/j.1582-4934.2008.00402.x.

Gorvel, Jean-Pierre, Philippe Chavrier, Marino Zerial, and Jean Gruenberg. 1991. “Rab5 Controls Early Endosome Fusion in Vitro.” Cell 64 (5): 915–25. https://doi.org/10.1016/0092-8674(91)90316-Q.

Goto-Silva, Livia, Marisa P McShane, Sara Salinas, Yannis Kalaidzidis, Giampietro Schiavo,

and Marino Zerial. 2019. “Retrograde Transport of Akt by a Neuronal Rab5-APPL1 Endosome.” Scientific Reports 9 (1): 2433. https://doi.org/10.1038/s41598-019-38637-0.

Gramlich, Michael W, and Vitaly A Klyachko. 2017. “Actin/Myosin-V- and Activity-Dependent Inter-Synaptic Vesicle Exchange in Central Neurons.” Cell Reports 18 (9): 2096–2104. https://doi.org/10.1016/j.celrep.2017.02.010.

Guo, Chunyan, Li Sun, Xueping Chen, and Danshen Zhang. 2013. “Oxidative Stress, Mitochondrial Damage and Neurodegenerative Diseases.” Neural Regeneration Research 8 (21): 2003–14. https://doi.org/10.3969/j.issn.1673-5374.2013.21.009.

Haddad, Dominik M., Sven Vilain, Melissa Vos, Giovanni Esposito, Samer Matta, Vera M. Kalscheuer, Katleen Craessaerts, et al. 2013. “Mutations in the Intellectual Disability Gene Ube2a Cause Neuronal Dysfunction and Impair Parkin-Dependent Mitophagy.” Molecular Cell. https://doi.org/10.1016/j.molcel.2013.04.012.

Hall, Emma A, Michael S Nahorski, Lyndsay M Murray, Ranad Shaheen, Emma Perkins, Kosala N Dissanayake, Yosua Kristaryanto, et al. 2017. “PLAA Mutations Cause a Lethal Infantile Epileptic Encephalopathy by Disrupting Ubiquitin-Mediated Endolysosomal Degradation of Synaptic Proteins.” American Journal of Human Genetics 100 (5): 706–24. https://doi.org/10.1016/j.ajhg.2017.03.008.

Hayakawa, Akira, and Naomi Kitamura. 2000. “Early Endosomal Localization of Hrs Requires a Sequence within the Proline- and Glutamine-Rich Region but Not the FYVE Finger.” Journal of Biological Chemistry 275 (38): 29636–42. https://doi.org/10.1074/jbc.M002696200.

Hendricks, Adam G, Eran Perlson, Jennifer L Ross, Harry W Schroeder III, Mariko Tokito, and Erika L F Holzbaur. 2010. “Motor Coordination via a Tug-of-War Mechanism Drives Bidirectional Vesicle Transport.” Current Biology 20 (8): 697–702. https://doi.org/10.1016/j.cub.2010.02.058.

Henne, William M., Nicholas J. Buchkovich, and Scott D. Emr. 2011. “The ESCRT Pathway.” Developmental Cell 21 (1): 77–91. https://doi.org/10.1016/j.devcel.2011.05.015.

Hernandez, Daniela, Ciara A. Torres, Wanda Setlik, Carolina Cebrián, Eugene V. Mosharov, Guomei Tang, Hsiao Chun Cheng, et al. 2012. “Regulation of Presynaptic Neurotransmission by Macroautophagy.” Neuron. https://doi.org/10.1016/j.neuron.2012.02.020.

Huckaba, Thomas M, Arne Gennerich, James E Wilhelm, Athar H Chishti, and Ronald D Vale. 2011. “Kinesin-73 Is a Processive Motor That Localizes to Rab5-Containing Organelles.” Journal of Biological Chemistry 286 (9): 7457–67. https://doi.org/10.1074/jbc.M110.167023.

Hurley, James H. 2015. “ESCRTs Are Everywhere.” The EMBO Journal 34 (19): 2398–2407.

https://doi.org/10.15252/embj.201592484.

Husnjak, Koraljka, and Ivan Dikic. 2012. “Ubiquitin-Binding Proteins: Decoders of Ubiquitin- Mediated Cellular Functions.” Annual Review of Biochemistry 81 (1): 291–322. https://doi.org/10.1146/annurev-biochem-051810-094654.

Im, Young Jun, and James H. Hurley. 2008. “Integrated Structural Model and Membrane Targeting Mechanism of the Human ESCRT-II Complex.” Developmental Cell. https://doi.org/10.1016/j.devcel.2008.04.004.

Jenkins, Brian, Helena Decker, Marvin Bentley, Julie Luisi, and Gary Banker. 2012. “A Novel Split Kinesin Assay Identifies Motor Proteins That Interact with Distinct Vesicle Populations.” Journal of Cell Biology 198 (4): 749–61. https://doi.org/10.1083/jcb.201205070.

Kabeya, Y. 2000. “LC3, a Mammalian Homologue of Yeast Apg8p, Is Localized in Autophagosome Membranes after Processing.” The EMBO Journal. https://doi.org/10.1093/emboj/19.21.5720.

Kaech, Stefanie, and Gary Banker. 2006. “Culturing Hippocampal Neurons.” Nature Protocols 1 (5): 2406–15. https://doi.org/10.1038/nprot.2006.356.

Katzmann, David J., Markus Babst, and Scott D. Emr. 2001. “Ubiquitin-Dependent Sorting into the Multivesicular Body Pathway Requires the Function of a Conserved Endosomal Protein Sorting Complex, ESCRT-I.” Cell. https://doi.org/10.1016/S0092-8674(01)00434-2.

Komada, Masayuki, and Philippe Soriano. 1999. “Hrs, a FYVE Finger Protein Localized to Early Endosomes, Is Implicated in Vesicular Traffic and Required for Ventral Folding Morphogenesis.” Genes & Development 13 (11): 1475–85. http://genesdev.cshlp.org/content/13/11/1475.abstract.

Kononenko, Natalia L., Gala A. Claßen, Marijn Kuijpers, Dmytro Puchkov, Tanja Maritzen, Aleksandra Tempes, Anna R. Malik, et al. 2017. “Retrograde Transport of TrkB-Containing Autophagosomes via the Adaptor AP-2 Mediates Neuronal Complexity and Prevents Neurodegeneration.” Nature Communications. https://doi.org/10.1038/ncomms14819.

Kulkarni, Vineet Vinay, and Sandra Maday. 2018a. “Compartment-Specific Dynamics and Functions of Autophagy in Neurons.” Developmental Neurobiology 78 (3): 298–310. https://doi.org/10.1002/dneu.22562.

———. 2018b. “Neuronal Endosomes to Lysosomes: A Journey to the Soma.” Journal of Cell Biology 217 (9): 2977–79. https://doi.org/10.1083/jcb.201806139.

Kwong, J, F L Roundabush, P Hutton Moore, M Montague, W Oldham, Y Li, L S Chin, and L Li. 2000. “Hrs Interacts with SNAP-25 and Regulates Ca(2+)-Dependent Exocytosis.” Journal of Cell Science 113 (12): 2273 LP – 2284.

http://jcs.biologists.org/content/113/12/2273.abstract.

Langemeyer, Lars, Florian Fröhlich, and Christian Ungermann. 2018. “Rab GTPase Function in Endosome and Lysosome Biogenesis.” Trends in Cell Biology 28 (11): 957–70. https://doi.org/10.1016/j.tcb.2018.06.007.

Lawe, Deirdre C, Varsha Patki, Robin Heller-Harrison, David Lambright, and Silvia Corvera. 2000. “The FYVE Domain of Early Endosome Antigen 1 Is Required for Both Phosphatidylinositol 3-Phosphate and Rab5 Binding: CRITICAL ROLE OF THIS DUAL INTERACTION FOR ENDOSOMAL LOCALIZATION .” Journal of Biological Chemistry 275 (5): 3699–3705. https://doi.org/10.1074/jbc.275.5.3699.

Leal-Ortiz, Sergio, Clarissa L Waites, Ryan Terry-Lorenzo, Pedro Zamorano, Eckart D Gundelfinger, and Craig C Garner. 2008. “Piccolo Modulation of Synapsin1a Dynamics Regulates Synaptic Vesicle Exocytosis .” The Journal of Cell Biology 181 (5): 831–46. https://doi.org/10.1083/jcb.200711167.

Lee, DaWon, Xiaoyue Zheng, Kay Shigemori, Christopher Krasniak, Jie Bin Liu, Chao Tang, Joshua Kavaler, and S Tariq Ahmad. 2019. “Expression of Mutant CHMP2B Linked to Neurodegeneration in Humans Disrupts Circadian Rhythms in Drosophila.” FASEB BioAdvances 1 (8): 511–20. https://doi.org/10.1096/fba.2019-00042.

Lee, Jin- A, Anne Beigneux, S Tariq Ahmad, Stephen G Young, and Fen-Biao Gao. 2007. “ESCRT-III Dysfunction Causes Autophagosome Accumulation and Neurodegeneration.” Current Biology 17 (18): 1561–67. https://doi.org/http://doi.org/10.1016/j.cub.2007.07.029.

Lee, Jin A., Anne Beigneux, S. Tariq Ahmad, Stephen G. Young, and Fen Biao Gao. 2007. “ESCRT-III Dysfunction Causes Autophagosome Accumulation and Neurodegeneration.” Current Biology. https://doi.org/10.1016/j.cub.2007.07.029.

Lee, Juliette J, Alvaro Sanchez-Martinez, Aitor Martinez Zarate, Cristiane Benincá, Ugo Mayor, Michael J Clague, and Alexander J Whitworth. 2018. “Basal Mitophagy Is Widespread in Drosophila but Minimally Affected by Loss of Pink1 or Parkin.” The Journal of Cell Biology 217 (5): 1613. http://jcb.rupress.org/content/217/5/1613.abstract.

Leil, Tarek A, Zi-Wei Chen, Chang-Sheng S Chang, and Richard W Olsen. 2004. “GABAA Receptor-Associated Protein Traffics GABAA Receptors to the Plasma Membrane in Neurons.” The Journal of Neuroscience 24 (50): 11429. http://www.jneurosci.org/content/24/50/11429.abstract.

Leonard, Deborah, Akira Hayakawa, Deirdre Lawe, David Lambright, Karl D Bellve, Clive Standley, Lawrence M Lifshitz, Kevin E Fogarty, and Silvia Corvera. 2008. “Sorting of EGF and Transferrin at the Plasma Membrane and by Cargo-Specific Signaling to EEA1- Enriched Endosomes.” Journal of Cell Science 121 (20): 3445.

http://jcs.biologists.org/content/121/20/3445.abstract.

Lin, Mei-Yao, and Zu-Hang Sheng. 2015. “Regulation of Mitochondrial Transport in Neurons.” Experimental Cell Research 334 (1): 35–44. https://doi.org/https://doi.org/10.1016/j.yexcr.2015.01.004.

Lin, Mei Yao, Xiu Tang Cheng, Prasad Tammineni, Yuxiang Xie, Bing Zhou, Qian Cai, and Zu Hang Sheng. 2017. “Releasing Syntaphilin Removes Stressed Mitochondria from Axons Independent of Mitophagy under Pathophysiological Conditions.” Neuron. https://doi.org/10.1016/j.neuron.2017.04.004.

Liou, Willisa, Hans J Geuze, Math J H Geelen, and Jan W Slot. 1997. “The Autophagic and Endocytic Pathways Converge at the Nascent Autophagic Vacuoles.” The Journal of Cell Biology 136 (1): 61. http://jcb.rupress.org/content/136/1/61.abstract.

Maday, Sandra, and Erika L F Holzbaur. 2016. “Compartment-Specific Regulation of Autophagy in Primary Neurons.” The Journal of Neuroscience 36 (22): 5933. http://www.jneurosci.org/content/36/22/5933.abstract.

Maday, Sandra, Alison E. Twelvetrees, Armen J. Moughamian, and Erika L.F. Holzbaur. 2014. “Axonal Transport: Cargo-Specific Mechanisms of Motility and Regulation.” Neuron 84 (2): 292–309. https://doi.org/10.1016/j.neuron.2014.10.019.

Maday, Sandra, Karen E Wallace, and Erika L F Holzbaur. 2012. “Autophagosomes Initiate Distally and Mature during Transport toward the Cell Soma in Primary Neurons.” Journal of Cell Biology 196 (4): 407–17. https://doi.org/10.1083/jcb.201106120.

Mann, S. S., and J. A. Hammarback. 1994. “Molecular Characterization of Light Chain 3. A Microtubule Binding Subunit of MAP1A and MAP1B.” Journal of Biological Chemistry.

Matsuda, Noriyuki, Shigeto Sato, Kahori Shiba, Kei Okatsu, Keiko Saisho, Clement A. Gautier, Yu Shin Sou, et al. 2010. “PINK1 Stabilized by Mitochondrial Depolarization Recruits Parkin to Damaged Mitochondria and Activates Latent Parkin for Mitophagy.” Journal of Cell Biology. https://doi.org/10.1083/jcb.200910140.

Matsuda, Shinji, Eriko Miura, Keiko Matsuda, Wataru Kakegawa, Kazuhisa Kohda, Masahiko Watanabe, and Michisuke Yuzaki. 2008. “Accumulation of AMPA Receptors in Autophagosomes in Neuronal Axons Lacking Adaptor Protein AP-4.” Neuron. https://doi.org/10.1016/j.neuron.2008.02.012.

Mayers, Jonathan R, Ian Fyfe, Amber L Schuh, Edwin R Chapman, J Michael Edwardson, and Anjon Audhya. 2011. “ESCRT-0 Assembles as a Heterotetrameric Complex on Membranes and Binds Multiple Ubiquitinylated Cargoes Simultaneously.” Journal of Biological Chemistry 286 (11): 9636–45. https://doi.org/10.1074/jbc.M110.185363.

McCullough, John, Michael J. Clague, and Sylvie Urbé. 2004. “AMSH Is an Endosome-

Associated Ubiquitin Isopeptidase.” Journal of Cell Biology. https://doi.org/10.1083/jcb.200401141.

McMahon, Harvey T, and Thomas C Südhof. 1995. “Synaptic Core Complex of Synaptobrevin, Syntaxin, and SNAP25 Forms High Affinity -SNAP Binding Site .” Journal of Biological Chemistry 270 (5): 2213–17. https://doi.org/10.1074/jbc.270.5.2213.

Mizuno, Emi, Takanobu Iura, Akiko Mukai, Tamotsu Yoshimori, Naomi Kitamura, and Masayuki Komada. 2005. “Regulation of Epidermal Growth Factor Receptor Down- Regulation by UBPY-Mediated Deubiquitination at Endosomes.” Molecular Biology of the Cell. https://doi.org/10.1091/mbc.E05-06-0560.

Mohrlüder, Jeannine, Melanie Schwarten, and Dieter Willbold. 2009. “Structure and Potential Function of Γ‐aminobutyrate Type A Receptor‐associated Protein.” The FEBS Journal 276 (18): 4989–5005. https://doi.org/10.1111/j.1742-4658.2009.07207.x.

Montanara, Paolo La, Laura Rusconi, Albina Locarno, Lia Forti, Isabella Barbiero, Marco Tramarin, Chetan Chandola, Charlotte Kilstrup-Nielsen, and Nicoletta Landsberger. 2015. “Synaptic Synthesis, Dephosphorylation, and Degradation: A NOVEL PARADIGM FOR AN ACTIVITY-DEPENDENT NEURONAL CONTROL OF CDKL5 .” Journal of Biological Chemistry 290 (7): 4512–27. https://doi.org/10.1074/jbc.M114.589762.

Mouton-Liger, Francois, Maxime Jacoupy, Jean-Christophe Corvol, and Olga Corti. 2017. “PINK1/Parkin-Dependent Mitochondrial Surveillance: From Pleiotropy to Parkinson’s Disease .” Frontiers in Molecular Neuroscience . https://www.frontiersin.org/article/10.3389/fnmol.2017.00120.

Muzioł, Tadeusz, Estela Pineda-Molina, Raimond B. Ravelli, Alessia Zamborlini, Yoshiko Usami, Heinrich Göttlinger, and Winfried Weissenhorn. 2006. “Structural Basis for Budding by the ESCRT-III Factor CHMP3.” Developmental Cell. https://doi.org/10.1016/j.devcel.2006.03.013.

Narendra, Derek P., Seok Min Jin, Atsushi Tanaka, Der Fen Suen, Clement A. Gautier, Jie Shen, Mark R. Cookson, and Richard J. Youle. 2010. “PINK1 Is Selectively Stabilized on Impaired Mitochondria to Activate Parkin.” PLoS Biology. https://doi.org/10.1371/journal.pbio.1000298.

Obashi, Kazuki, and Shigeo Okabe. 2013. “Regulation of Mitochondrial Dynamics and Distribution by Synapse Position and Neuronal Activity in the Axon.” European Journal of Neuroscience 38 (3): 2350–63. https://doi.org/10.1111/ejn.12263.

Ohtake, Fumiaki, Hikaru Tsuchiya, Yasushi Saeki, and Keiji Tanaka. 2018. “K63 Ubiquitylation Triggers Proteasomal Degradation by Seeding Branched Ubiquitin Chains.” Proceedings of the National Academy of Sciences 115 (7): E1401 LP-E1408. https://doi.org/10.1073/pnas.1716673115.

Okerlund, Nathan D., Katharina Schneider, Sergio Leal-Ortiz, Carolina Montenegro-Venegas, Sally A. Kim, Loren C. Garner, Clarissa L. Waites, Eckart D. Gundelfinger, Richard J. Reimer, and Craig C. Garner. 2018. “Bassoon Controls Presynaptic Autophagy through Atg5.” Neuron. https://doi.org/10.1016/j.neuron.2018.01.010.

Olenick, Mara A, Roberto Dominguez, and Erika L F Holzbaur. 2018. “Dynein Activator Hook1 Is Required for Trafficking of BDNF-Signaling Endosomes in Neurons.” The Journal of Cell Biology 218 (1): 220–33. https://doi.org/10.1083/jcb.201805016.

Parkinson, N., P. G. Ince, M. O. Smith, R. Highley, G. Skibinski, P. M. Andersen, K. E. Morrison, et al. 2006. “ALS Phenotypes with Mutations in CHMP2B (Charged Multivesicular Body Protein 2B).” Neurology. https://doi.org/10.1212/01.wnl.0000231510.89311.8b.

Poteryaev, Dmitry, Sunando Datta, Karin Ackema, Marino Zerial, and Anne Spang. 2010. “Identification of the Switch in Early-to-Late Endosome Transition.” Cell 141 (3): 497–508. https://doi.org/10.1016/j.cell.2010.03.011.

Qu, Xiaoyi, Atul Kumar, Heike Blockus, Clarissa Waites, and Francesca Bartolini. 2019. “Activity-Dependent Nucleation of Dynamic Microtubules at Presynaptic Boutons Controls Neurotransmission.” Current Biology 29 (24): 4231-4240.e5. https://doi.org/10.1016/j.cub.2019.10.049.

Raiborg, Camilla, Bjørn Bremnes, Anja Mehlum, David J Gillooly, Antonello D’Arrigo, Espen Stang, and Harald Stenmark. 2001. “FYVE and Coiled-Coil Domains Determine the Specific Localisation of Hrs to Early Endosomes.” Journal of Cell Science 114 (12): 2255 LP – 2263. http://jcs.biologists.org/content/114/12/2255.abstract.

Raiborg, Camilla, Tor Erik Rusten, and Harald Stenmark. 2003. “Protein Sorting into Multivesicular Endosomes.” Current Opinion in Cell Biology 15 (4): 446–55. https://doi.org/https://doi.org/10.1016/S0955-0674(03)00080-2.

Ren, Xuefeng, and James H. Hurley. 2010. “VHS Domains of ESCRT-0 Cooperate in High- Avidity Binding to Polyubiquitinated Cargo.” EMBO Journal. https://doi.org/10.1038/emboj.2010.6.

Rodriguez-Enriquez, Sara, Insil Kim, Robert T. Currin, and John J. Lemasters. 2006. “Tracker Dyes to Probe Mitochondrial Autophagy (Mitophagy) in Rat Hepatocytes.” Autophagy. https://doi.org/10.4161/auto.2229.

Row, Paula E., Ian A. Prior, John McCullough, Michael J. Clague, and Sylvie Urbé. 2006. “The Ubiquitin Isopeptidase UBPY Regulates Endosomal Ubiquitin Dynamics and Is Essential for Receptor Down-Regulation.” Journal of Biological Chemistry. https://doi.org/10.1074/jbc.M512615200.

Rowland, Aaron M., Janet E. Richmond, Jason G. Olsen, David H. Hall, and Bruce A. Bamber.

2006. “Presynaptic Terminals Independently Regulate Synaptic Clustering and Autophagy of GABAA Receptors in Caenorhabditis Elegans.” Journal of Neuroscience. https://doi.org/10.1523/JNEUROSCI.2279-05.2006.

Rusten, T E, and Harald Stenmark. 2009. “How Do ESCRT Proteins Control Autophagy?” Journal of Cell Science 122 (13): 2179 LP – 2183. https://doi.org/10.1242/jcs.050021.

Rusten, Tor Erik, and Anne Simonsen. 2008. “ESCRT Functions in Autophagy and Associated Disease.” Cell Cycle 7 (9): 1166–72. https://doi.org/10.4161/cc.7.9.5784.

Schu, P V, K Takegawa, M J Fry, J H Stack, M D Waterfield, and S D Emr. 1993. “Phosphatidylinositol 3-Kinase Encoded by Yeast VPS34 Gene Essential for Protein Sorting.” Science 260 (5104): 88 LP – 91. https://doi.org/10.1126/science.8385367.

Schuh, Amber L, and Anjon Audhya. 2014. “The ESCRT Machinery: From the Plasma Membrane to Endosomes and Back Again.” Critical Reviews in Biochemistry and Molecular Biology 49 (3): 242–61. https://doi.org/10.3109/10409238.2014.881777.

Schwarz, Lindsay A., Benjamin J. Hall, and Gentry N. Patrick. 2010. “Activity-Dependent Ubiquitination of GluA1 Mediates a Distinct AMPA Receptor Endocytosis and Sorting Pathway.” Journal of Neuroscience. https://doi.org/10.1523/JNEUROSCI.3686-10.2010.

Sheehan, Patricia, Mei Zhu, Anne Beskow, Cyndel Vollmer, and Clarissa L Waites. 2016. “Activity-Dependent Degradation of Synaptic Vesicle Proteins Requires Rab35 and the ESCRT Pathway.” The Journal of Neuroscience 36 (33): 8668. http://www.jneurosci.org/content/36/33/8668.abstract.

Shehata, Mohammad, Hiroyuki Matsumura, Reiko Okubo-Suzuki, Noriaki Ohkawa, and Kaoru Inokuchi. 2012. “Neuronal Stimulation Induces Autophagy in Hippocampal Neurons That Is Involved in AMPA Receptor Degradation after Chemical Long-Term Depression.” The Journal of Neuroscience 32 (30): 10413 LP – 10422. https://doi.org/10.1523/JNEUROSCI.4533-11.2012.

Shen, Wei, and Barry Ganetzky. 2009. “Autophagy Promotes Synapse Development in Drosophila.” Journal of Cell Biology. https://doi.org/10.1083/jcb.200907109.

Shenoy, Sudha K., and Robert J. Lefkowitz. 2003. “Trafficking Patterns of β-Arrestin and G Protein-Coupled Receptors Determined by the Kinetics of β-Arrestin Deubiquitination.” Journal of Biological Chemistry. https://doi.org/10.1074/jbc.M209626200.

Shimizu, Hideyuki, Satoru Kawamura, and Koichi Ozaki. 2003. “An Essential Role of Rab5 in Uniformity of Synaptic Vesicle Size.” Journal of Cell Science 116 (17): 3583 LP – 3590. https://doi.org/10.1242/jcs.00676.

Simonsen, Anne, Roger Lippe, Savvas Christoforidis, Jean-Michel Gaullier, Andreas Brech, Judy Callaghan, Ban-Hock Toh, Carol Murphy, Marino Zerial, and Harald Stenmark. 1998.

“EEA1 Links PI(3)K Function to Rab5 Regulation of Endosome Fusion.” Nature 394 (6692): 494–98. https://doi.org/10.1038/28879.

Skibinski, Gaia, Nicholas J. Parkinson, Jeremy M. Brown, Lisa Chakrabarti, Sarah L. Lloyd, Holger Hummerich, Jørgen E. Nielsen, et al. 2005. “Mutations in the Endosomal ESCRTIII-Complex Subunit CHMP2B in Frontotemporal Dementia.” Nature Genetics. https://doi.org/10.1038/ng1609.

Söllner, Thomas, Sidney W Whiteheart, Michael Brunner, Hediye Erdjument-Bromage, Scott Geromanos, Paul Tempst, and James E Rothman. 1993. “SNAP Receptors Implicated in Vesicle Targeting and Fusion.” Nature 362 (6418): 318–24. https://doi.org/10.1038/362318a0.

Song, Ai-hong, Dong Wang, Gang Chen, Yuju Li, Jianhong Luo, Shumin Duan, and Mu-ming Poo. 2009. “A Selective Filter for Cytoplasmic Transport at the Axon Initial Segment.” Cell 136 (6): 1148–60. https://doi.org/https://doi.org/10.1016/j.cell.2009.01.016.

Soukup, Sandra-Fausia, Sabine Kuenen, Roeland Vanhauwaert, Julia Manetsberger, Sergio Hernández-Díaz, Jef Swerts, Nils Schoovaerts, et al. 2016. “A LRRK2-Dependent EndophilinA Phosphoswitch Is Critical for Macroautophagy at Presynaptic Terminals.” Neuron 92 (4): 829–44. https://doi.org/https://doi.org/10.1016/j.neuron.2016.09.037.

Spencer, Brian, Sharareh Emadi, Paula Desplats, Simona Eleuteri, Sarah Michael, Kori Kosberg, Jay Shen, et al. 2014. “ESCRT-Mediated Uptake and Degradation of Brain-Targeted α-Synuclein Single Chain Antibody Attenuates Neuronal Degeneration In Vivo.” Molecular Therapy 22 (10): 1753–67. https://doi.org/10.1038/mt.2014.129.

Spencer, Brian, Changyoun Kim, Tania Gonzalez, Alejandro Bisquertt, Christina Patrick, Edward Rockenstein, Anthony Adame, Seung-Jae Lee, Paula Desplats, and Eliezer Masliah. 2016. “α-Synuclein Interferes with the ESCRT-III Complex Contributing to the Pathogenesis of Lewy Body Disease.” Human Molecular Genetics 25 (6): 1100–1115. https://doi.org/10.1093/hmg/ddv633.

Stahelin, Robert V, Fei Long, Karthikeyan Diraviyam, Karol S Bruzik, Diana Murray, and Wonhwa Cho. 2002. “Phosphatidylinositol 3-Phosphate Induces the Membrane Penetration of the FYVE Domains of Vps27p and Hrs.” Journal of Biological Chemistry 277 (29): 26379–88. https://doi.org/10.1074/jbc.M201106200.

Stavoe, Andrea K.H., Sarah E. Hill, David H. Hall, and Daniel A. Colón-Ramos. 2016. “KIF1A/UNC-104 Transports ATG-9 to Regulate Neurodevelopment and Autophagy at Synapses.” Developmental Cell 38 (2): 171–85. https://doi.org/10.1016/j.devcel.2016.06.012.

Sun, Wei, Qing Yan, Thomas A Vida, and Andrew J Bean. 2003. “Hrs Regulates Early Endosome Fusion by Inhibiting Formation of an Endosomal SNARE Complex .” The Journal of Cell Biology 162 (1): 125–37. https://doi.org/10.1083/jcb.200302083.

Sung, Hyun, Lauren C. Tandarich, Kenny Nguyen, and Peter J. Hollenbeck. 2016. “Compartmentalized Regulation of Parkin-Mediated Mitochondrial Quality Control in the Drosophila Nervous System in Vivo.” Journal of Neuroscience. https://doi.org/10.1523/JNEUROSCI.0633-16.2016.

Takahashi, Hirohide, Jonathan R Mayers, Lei Wang, J Michael Edwardson, and Anjon Audhya. 2015. “Hrs and STAM Function Synergistically to Bind Ubiquitin-Modified Cargoes in Vitro.” Biophysical Journal 108 (1): 76–84. https://doi.org/10.1016/j.bpj.2014.11.004.

Tamai, Keiichi, Masafumi Toyoshima, Nobuyuki Tanaka, Noriko Yamamoto, Yuji Owada, Hiroshi Kiyonari, Kazuko Murata, et al. 2008. “Loss of Hrs in the Central Nervous System Causes Accumulation of Ubiquitinated Proteins and Neurodegeneration.” The American Journal of Pathology 173 (6): 1806–17. https://doi.org/http://doi.org/10.2353/ajpath.2008.080684.

Tang, Guomei, Kathryn Gudsnuk, Sheng Han Kuo, Marisa L. Cotrina, Gorazd Rosoklija, Alexander Sosunov, Mark S. Sonders, et al. 2014. “Loss of MTOR-Dependent Macroautophagy Causes Autistic-like Synaptic Pruning Deficits.” Neuron. https://doi.org/10.1016/j.neuron.2014.07.040.

Tanida, Isei, Takashi Ueno, and Eiki Kominami. 2004. “LC3 Conjugation System in Mammalian Autophagy.” The International Journal of Biochemistry & Cell Biology 36 (12): 2503–18. https://doi.org/https://doi.org/10.1016/j.biocel.2004.05.009.

Taoufik, Era, Georgia Kouroupi, Ourania Zygogianni, and Rebecca Matsas. 2020. “Synaptic Dysfunction in Neurodegenerative and Neurodevelopmental Diseases: An Overview of Induced Pluripotent Stem-Cell-Based Disease Models.” Open Biology 8 (9): 180138. https://doi.org/10.1098/rsob.180138.

Truckenbrodt, Sven, Abhiyan Viplav, Sebastian Jähne, Angela Vogts, Annette Denker, Hanna Wildhagen, Eugenio F Fornasiero, and Silvio O Rizzoli. 2018. “Newly Produced Synaptic Vesicle Proteins Are Preferentially Used in Synaptic Transmission.” The EMBO Journal 37 (15): e98044. https://doi.org/10.15252/embj.201798044.

Uytterhoeven, Valerie, Sabine Kuenen, Jaroslaw Kasprowicz, Katarzyna Miskiewicz, and Patrik Verstreken. 2011. “Loss of Skywalker Reveals Synaptic Endosomes as Sorting Stations for Synaptic Vesicle Proteins.” Cell 145 (1): 117–32. https://doi.org/10.1016/j.cell.2011.02.039.

———. 2017. “Loss of Skywalker Reveals Synaptic Endosomes as Sorting Stations for Synaptic Vesicle Proteins.” Cell 145 (1): 117–32. https://doi.org/10.1016/j.cell.2011.02.039.

Vagnozzi, Alana N, and Domenico Praticò. 2019. “Endosomal Sorting and Trafficking, the Retromer Complex and Neurodegeneration.” Molecular Psychiatry 24 (6): 857–68. https://doi.org/10.1038/s41380-018-0221-3.

Vanhauwaert, Roeland, Sabine Kuenen, Roy Masius, Adekunle Bademosi, Julia Manetsberger, Nils Schoovaerts, Laura Bounti, et al. 2017. “The SAC1 Domain in Synaptojanin Is Required for Autophagosome Maturation at Presynaptic Terminals.” The EMBO Journal 36 (10): 1392–1411. https://doi.org/10.15252/embj.201695773.

Vaz-Silva, João, Patrícia Gomes, Qi Jin, Mei Zhu, Viktoriya Zhuravleva, Sebastian Quintremil, Torcato Meira, et al. 2018. “Endolysosomal Degradation of Tau and Its Role in Glucocorticoid-Driven Hippocampal Malfunction.” The EMBO Journal 37 (20): e99084. https://doi.org/10.15252/embj.201899084.

Vijayan, Vinoy, and Patrik Verstreken. 2017. “Autophagy in the Presynaptic Compartment in Health and Disease.” Journal of Cell Biology 216 (7): 1895–1906. https://doi.org/10.1083/jcb.201611113.

Vives-Bauza, Cristofol, Chun Zhou, Yong Huang, Mei Cui, Rosa L.A. De Vries, Jiho Kim, Jessica May, et al. 2010. “PINK1-Dependent Recruitment of Parkin to Mitochondria in Mitophagy.” Proceedings of the National Academy of Sciences of the United States of America. https://doi.org/10.1073/pnas.0911187107.

Waites, Clarissa L, Sergio A Leal-Ortiz, Nathan Okerlund, Hannah Dalke, Anna Fejtova, Wilko D Altrock, Eckart D Gundelfinger, and Craig C Garner. 2013. “Bassoon and Piccolo Maintain Synapse Integrity by Regulating Protein Ubiquitination and Degradation.” The EMBO Journal 32 (7): 954–69. https://doi.org/10.1038/emboj.2013.27.

Waites, Clarissa L, Christian G Specht, Kai Härtel, Sergio Leal-Ortiz, David Genoux, Dong Li, Renaldo C Drisdel, et al. 2009. “Synaptic SAP97 Isoforms Regulate AMPA Receptor Dynamics and Access to Presynaptic Glutamate.” The Journal of Neuroscience : The Official Journal of the Society for Neuroscience 29 (14): 4332–45. https://doi.org/10.1523/JNEUROSCI.4431-08.2009.

Wang, Tong, Sally Martin, Andreas Papadopulos, Callista B Harper, Timur A Mavlyutov, Dhevahi Niranjan, Nick R Glass, et al. 2015. “Control of Autophagosome Axonal Retrograde Flux by Presynaptic Activity Unveiled Using Botulinum Neurotoxin Type A.” The Journal of Neuroscience 35 (15): 6179 LP – 6194. https://doi.org/10.1523/JNEUROSCI.3757-14.2015.

Wang, Xinnan, Dominic Winter, Ghazaleh Ashrafi, Julia Schlehe, Yao Liang Wong, Dennis Selkoe, Sarah Rice, Judith Steen, Matthew J. Lavoie, and Thomas L. Schwarz. 2011. “PINK1 and Parkin Target Miro for Phosphorylation and Degradation to Arrest Mitochondrial Motility.” Cell. https://doi.org/10.1016/j.cell.2011.10.018.

Wenzel, Eva Maria, Sebastian Wolfgang Schultz, Kay Oliver Schink, Nina Marie Pedersen, Viola Nähse, Andreas Carlson, Andreas Brech, Harald Stenmark, and Camilla Raiborg. 2018. “Concerted ESCRT and Clathrin Recruitment Waves Define the Timing and Morphology of Intraluminal Vesicle Formation.” Nature Communications 9 (1): 2932. https://doi.org/10.1038/s41467-018-05345-8.

Wilson, Jean M, Meltsje de Hoop, Natasha Zorzi, Ban-Hock Toh, Carlos G Dotti, and Robert G Parton. 2000. “EEA1, a Tethering Protein of the Early Sorting Endosome, Shows a Polarized Distribution in Hippocampal Neurons, Epithelial Cells, and Fibroblasts.” Molecular Biology of the Cell 11 (8): 2657–71. https://doi.org/10.1091/mbc.11.8.2657.

Yap, Chan Choo, Laura Digilio, Lloyd P McMahon, A Denise R Garcia, and Bettina Winckler. 2018. “Degradation of Dendritic Cargos Requires Rab7-Dependent Transport to Somatic Lysosomes.” Journal of Cell Biology 217 (9): 3141–59. https://doi.org/10.1083/jcb.201711039.

Zhang, Kai, Rotem Fishel Ben Kenan, Yasuko Osakada, Wei Xu, Rachel S Sinit, Liang Chen, Xiaobei Zhao, Jia-Yun Chen, Bianxiao Cui, and Chengbiao Wu. 2013. “Defective Axonal Transport of Rab7 GTPase Results in Dysregulated Trophic Signaling.” The Journal of Neuroscience 33 (17): 7451 LP – 7462. https://doi.org/10.1523/JNEUROSCI.4322-12.2013.

Zhang, Yu, Benjamin Schmid, Nanett K. Nikolaisen, Mikkel A. Rasmussen, Blanca I. Aldana, Mikkel Agger, Kirstine Calloe, et al. 2017. “Patient IPSC-Derived Neurons for Disease Modeling of Frontotemporal Dementia with Mutation in CHMP2B.” Stem Cell Reports. https://doi.org/10.1016/j.stemcr.2017.01.012.

Zheng, Ning, Okunola Jeyifous, Charlotte Munro, Johanna M. Montgomery, and William N. Green. 2015. “Synaptic Activity Regulates AMPA Receptor Trafficking through Different Recycling Pathways.” ELife. https://doi.org/10.7554/eLife.06878.

Zhou, Ruyun, Shinsuke Niwa, Laurent Guillaud, Ying Tong, and Nobutaka Hirokawa. 2013. “A Molecular Motor, KIF13A, Controls Anxiety by Transporting the Serotonin Type 1A Receptor.” Cell Reports 3 (2): 509–19. https://doi.org/10.1016/j.celrep.2013.01.014.

100