Autophagosome biogenesis in primary neurons follows an ordered and spatially regulated pathway

Summary

Autophagy is an essential degradative pathway in neurons, yetlittle is known about the mechanisms driving this process in highly polarized cells. Here, we use dual-color live-cell imaging to investigate the neuron-specific mechanisms of constitutiveautophagosome biogenesis in primary DRG and hippocampal cultures. Under basal conditions autophagosomes are continuously generated in the axon tip. There is an ordered assembly of proteins recruited with stereotypical kinetics onto the developing autophagosome. Plasma- or mitochondrial-derived membraneswere not incorporated into nascent autophagosomes in the distal axon. Rather, autophagosomes are generated at DFCP1-positive subdomains of the endoplasmic reticulum, distinct from ER exit sites. Biogenesis events arehighly enriched distally; autophagosomes form infrequently in dendrites, the cell soma or mid-axon, consistent with a highly compartmentalized pathway for constitutive autophagy in primary neurons. This distal biogenesis may facilitate the degradation of damaged mitochondria and long-lived cytoplasmic proteins that reach the axon tip via slow axonal transport.

Introduction

Macroautophagy (autophagy)is an evolutionarily conserved lysosomal degradation pathway that maintains the homeostasis of the cellular environment by eliminating damaged organelles and aggregated proteins(Xie and Klionsky, 2007). This pathway is particularly important in post-mitotic cells such as neurons that are unableto dilute out proteotoxins by cell division. Neuron-specific knockout of genes required for autophagy induces neuron cell death, marked by the accumulation of ubiquitin-positive aggregates(Hara et al., 2006; Komatsu et al., 2006). Further, mutations in the mitophagy machinery PINK1 and Parkin lead to early onset Parkinson’s disease(Kitada et al., 1998; Narendra et al., 2008; Narendra et al., 2010; Valente et al., 2004), potentially linking defective autophagy with the progression of neurodegeneration. Autophagy is thus a protective mechanism against neuronal dysfunction and degeneration(Marino et al., 2011; Rubinsztein et al., 2005).

Despite the evidence that autophagy is critical in maintaining neuronal homeostasis, little is understood about the mechanisms driving this process in neurons. Much of the work dissecting the autophagic pathway has been performed in yeast and nonpolarized mammalian cells(Mizushima et al., 2011; Weidberg et al., 2011). However, neurons exhibit a highly polarized and elongated morphology that poses a unique challenge to cellular trafficking and transport pathways. Many neurodegenerative disease-associated mutations have been identified in the machinery that transports organelles and proteins across the extended distance of the axon(Millecamps and Julien, 2013; Perlson et al., 2010), emphasizing the unique vulnerability of the neuronal system. Further, the majority of studies to date have focused on stress-induced autophagy as a result of nutrient deprivation. However, knockout mouse models have demonstrated that basal levels of autophagy are essential for neuronal survival(Hara et al., 2006; Komatsu et al., 2006). Thus, we set out to determine the spatiotemporal dynamics of autophagy in primary neurons under basal growth conditions.

Here, we describe the assembly of an autophagosome in real-time in primary dorsal root ganglion (DRG) and hippocampal neurons. We find that autophagosome formation in primary neuronsis a constitutive process under basal growth conditions. Autophagosomes are continuously generated at the axon tip; this biogenesisinvolves an ordered recruitment of proteins that assemble onto the nascent autophagosome with stereotypical kinetics. Under basal conditions, we find that neuronal autophagosomes are not generated from plasma- or mitochondrial-derived membrane, but rather from specific subdomains of the endoplasmic reticulum. Most strikingly, autophagosomes are preferentially generated at the distal end of the axon, emphasizing the high degree of compartmentalization and spatial regulation controlling autophagosome biogenesis in primary neurons. We propose that this pathway provides a mechanism to recycle distally-localized aged mitochondria as well as long-lived proteins transported to the distal axon via slow axonal transport.

Results

Autophagosome biogenesis in primary neurons follows an ordered and spatially regulated pathway

Autophagy is initiated when components of the cytosol are enveloped within a membrane cisterna termed an isolation membrane or phagophore(Xie and Klionsky, 2007). The edges of the isolation membrane fuse to form a closed double membrane autophagosome. These autophagosomes are shuttled toward lysosomes for fusion, and the contents are degraded by lysosomal hydrolases and recycled back into the cytosol. Seminal work in yeast identified proteins important for the formation of an autophagosome(Harding et al., 1995; Klionsky et al., 2003; Thumm et al., 1994; Tsukada and Ohsumi, 1993)and many are conserved in mammals(Mizushima et al., 2011; Weidberg et al., 2011). During autophagosome formation, these proteins are recruited to the assembly site in an ordered fashion(Itakura and Mizushima, 2010; Suzuki et al., 2001; Suzuki et al., 2007). We set out to determine whether neurons employ this conserved mechanism for autophagosome formation and define the temporal relationship between assembly proteins as they arrive at the nascent autophagosome.

We performed live-cell imaging on DRGneurons isolated from transgenic mice expressing GFPLC3. To measure autophagosome assembly, neurons were transfected with either Atg13 or Atg5 labeled with mCherry to allow pair-wise imaging with GFP-LC3. Atg13 is a member of the first macromolecular complex that initiates autophagosome formation(Mizushima et al., 2011) and Atg5 localizes to the isolation membrane and regulates its elongation(Mizushima et al., 2001). Thus, both Atg13 and Atg5 identify the initial stages of autophagosome formation. In all experiments, only those neurons with low Atg13 or Atg5 expression were imaged in order to observe specific events of biogenesis.

In primary DRG neurons, we observed the appearance of puncta positive for Atg13 that grew progressively in size (Fig. 1A). These structures appeared almost exclusively in the distal end of the axon. Following Atg13 arrival, LC3 was recruited to nascent autophagosomes. Atg13 then dissociated,leaving LC3-positive puncta to grow progressively into ring structures ~800 nm in diameter. We observed the same pattern of ordered recruitment between Atg5 and LC3 (Fig. 1B and Movie S1). Atg5 appeared initially, followed by LC3. After a period of overlap, the Atg5 signal decayed from nascent autophagosomes while the LC3 puncta grew into ring structures (Fig. 1B and Movie S1). Occasionally, we observed LC3-positive cup-shaped structures representing the isolation membrane (Fig. S1). Kymograph analysis of single biogenesis eventsclearly reveals the sequential appearance of Atg5 followed by LC3 in the distal axon (Fig 1G). Together, these results indicate that neuronal autophagosomes are generated by an ordered recruitment of assembly factors.

Figure 1. Autophagosome biogenesis in DRGneurons involves an ordered recruitment of assembly factors.

(A)Time series of mCherry-Atg13 and GFP-LC3 in the distal axon of DRG neurons. Red and green arrowheads indicate the ordered appearance of Atg13 followed by LC3, respectively. Yellow arrowheads denote colocalization between Atg13 and LC3. The retrograde direction is toward the right. Scale bar, 1 μm. (B) Time series of mCherry-Atg5 and GFP-LC3 in the distal axon. Red and green arrowheads indicate the ordered appearance of Atg5 followed by LC3, respectively. Yellow arrowheads denote colocalization between Atg5 and LC3. The retrograde direction is toward the right. Scale bar, 1 μm. See also Movie S1. (C) Intensity profiles of mCherry-Atg13 and GFP-LC3 from five individual biogenesis events. Vertical dashed line denotes the half-maximum in LC3 intensity, which was used to align the traces. (D) Mean intensity profile of mCherry-Atg13 and GFP-LC3 (mean± SEM; n=5 biogenesis events from 5 neurons from 3 separate experiments). (E) Intensity profiles of mCherry-Atg5 and GFP-LC3 from five individual biogenesis events. (F) Mean intensity profile of mCherry-Atg5 and GFP-LC3 (mean± SEM; n=5 biogenesis events from 3 neurons from 3 separate experiments). (G) Kymograph showing the ordered recruitment of mCherry-Atg5 followed by GFP-LC3 to nascent autophagosomes in the distal axon. Horizontal bar, 1 min. Vertical bar, 1 μm. (H, I) Trajectories of two autophagosome biogenesis events. Tracks are color coded for time (first frame is red; last frame is blue). (J) Alignment of Atg13 and LC3 intensity profiles with Atg5 and LC3 intensity profiles. Vertical dashed line marks the LC3 1/2 maximum to which the traces were aligned.

To resolve the temporal relationship between the recruitment of assembly factors, we quantitated the fluorescence intensity of Atg13 or Atg5 relative to that of LC3 over time. As shown in Figure 1C-F, the hierarchical recruitment of assembly factors proceeds with highly consistent kinetics. On average, Atg13 and Atg5 reached a maximum intensity within ~1.3 and ~1.0min after initial appearance, respectively, and sustained that signal for ~50 s before decay was initiated(Fig. 1C-F). Alignment of Atg13 and Atg5 mean intensity profiles indicated that they are recruited to nascent autophagosomes with kinetics that cannot be distinguished at this time resolution (1 frame every 2 s; Fig. 1J). However, Atg5 decayed from the nascent autophagosome prior to Atg13 (Fig. 1J). This order of disassembly from the nascent autophagosome was also observed with direct pair-wise imaging of GFP-Atg13 and mCherry-Atg5 (data not shown). The dynamics of LC3 intensity changes over time measured across all experiments were extraordinarily consistent. Following a lag phase, LC3 was robustly recruited, rising from initial signal to maximum intensity within 1.5 min (Fig. 1C-F, J). Decay of Atg5 was always initiated prior to observation of maximum LC3 intensity at the developing autophagosome (Fig. 1F, J).

Next, weasked whether there is a relationship between the stage of autophagosome maturation and the mobility of the structure. We hypothesized that perhaps during the Atg5-positive phase, the autophagosome is confined due to tetheringto the membrane source. Upon release of the closed autophagosome (Atg5-negative phase), the compartment may become more motile. To examine this possibility, we plotted the tracks exhibited by each biogenesis event. For two out of four biogenesis events, developing autophagosomes moved within a confined region within thedistal process (representative example in Fig. 1H). For the remaining twoevents, developing autophagosomes exhibited two phases of motility. Initially during the Atg5-positive phase, the nascent autophagosome exhibited a period of confined motility which was followedby a period of more unrestricted motility during the LC3-dominant phase (representative example in Fig. 1I). This transition in motility maycoincide with fusion of the isolation membrane andrelease of the closed autophagosome. While this sequence of events was only clearly observed in a subset of biogenesis events tracked, our resolution may be limited due to imaging in only a single focal plane.

Our previous observations on GFP-LC3 dynamics in primary neurons suggestedthat autophagosomes preferentially initiate distally(Maday et al., 2012). To examine this question more rigorously, we measured the localization of autophagosome biogenesis within the neuron, determining the number of autophagosomes that form in the distal axon, mid-axon and cell soma by monitoring either Atg13 or Atg5 (Fig. 2A). Strikingly, constitutiveautophagosomes were generated almost exclusively in the distal end of the axon (Fig. 2B-E). Comparing equivalent cross-sectional areas, the mean rate of autophagosome formation in the distal axon was ~20-fold higher than observed along the mid-axon (Fig. 2B,D). In the distal axon, autophagosome biogenesis was observed in all neurons with rates ranging from 0.2-1.3 autophagosomes formedper min (Fig. 2C, E). In contrast, ~70-90% of neurons displayed no biogenesis along the mid-axon during our observation period,withthe remaining neurons exhibiting biogenesis rates ≤0.2 autophagosomes per min (Fig. 2C,E). In the cell body, rates of formation were more variable, potentially due to imaging in a single focal plane within the ~5 μm depth of the soma. 86% of neurons displayed no biogenesis events in the cell body (Fig. 2B-E). While LC3-positive autophagosomes are clearly present along the mid-axon and in the cell soma, we did not observe robust formation in either of these regions under basal conditions. Thus, constitutive autophagy occurs preferentially at the distal end of the axon in primary DRG neurons.

Figure 2. Autophagosome biogenesis in primary neurons is enriched in the distal axon of DRG neurons.

(A) Schematic of a DRG neuron. The boxed regions designate the cell soma, axon and distal tip where rates of autophagosome formation were measured. The distal tip encompassed the terminal ~50 μm of the axon. Autophagosome formation along the axon (500-1000 μm in length) was quantified within an ~80 μm observation window. The cell soma encompassed a diameter of ~25 μm. Despite the difference in length, biogenesis rates in the axon and distal tip were measured within equivalent cross-sectional areas. (B) Mean rates of autophagosome formation (measured with mCherry-Atg13 and GFPLC3) in the cell soma, axon and distal tip (mean± SEM; cell soma, n=7 neurons from 3 experiments; axon, n=11 axons from 8 neurons and 4 experiments; distal tip, n=14 tips from 12 neurons and 4 experiments;ns, not significant; *, p<0.05; ***, p<0.001;one-way ANOVA with Tukey’s post-hoc test). (C) Distribution of individual biogenesis rates in each domainas measured with mCherry-Atg13. (D) Mean rates of autophagosome formation (measured with mCherry-Atg5 and GFP-LC3) in the cell soma, axon and distal tip (mean± SEM; cell soma, n=7 neurons from 2 experiments; axon, n=16 axons from 13 neurons and 3 experiments; distal tip, n=16 tips from 16 neurons and 5 experiments; ns, not significant; ***, p<0.001; one-way ANOVA with Tukey’s post-hoc test). (E) Distribution of individual biogenesis rates in each domain as measured with mCherry-Atg5.

Autophagosomes also form distally and undergo retrograde transport in axons of synaptically-connected neurons

While DRG neurons are a well-established model systemfor neurite development, these cellsactively extend processes and do not form synapses as a monoculture in vitro. Thus, we investigated whether the dynamics observed in DRG neurons are also seen in aculture that becomes synaptically-connected. To explore this possibility, we measured autophagosome biogenesis and transport along the axonsof hippocampal neurons cultured up to 16 DIV.

At 7 DIV, hippocampal neurons exhibit well-defined axon and dendrite projections (Fig. 3A) as defined morphologicallyand confirmed byMAP2 and tau immunostaining (not shown). By 10 DIV,excitatory synapses are evident (Fig.S2). Autophagosomes were detected along the length of the axon and in the cell soma of hippocampal neurons (Fig. 3A). Live cell imaging of the distal axon revealed active autophagosome formation with the appearance of GFP-LC3-positive puncta that grew into ring structures (Fig. 3B). Quantitation of the rates of biogenesis showed an enrichment of autophagosome formation in the distal axon as compared with the mid-axon (Fig. 3B). Biogenesis rates in either the mid or distal axon did not change with age of the culture; similar rates were observed at either young (5 DIV) or synaptically-connected (10 DIV) stages of development (Fig. 3B); in fact, robust distal biogenesis continued to be observed in neurons two weeks in vitro (not shown). While autophagosome formation in the cell soma was occasionally observed, measurement of biogenesis rates in this region of the cell was limited by the depth of the soma.

Figure 3. Autophagosomes form in the distal axon and undergo retrograde transport toward the cell soma in hippocampal neurons.

(A) Hippocampal neurons grown 7 DIV form axon and dendrite processes. Arrowheads denote autophagosomes undergoing retrograde transport in the axon and entering the cell soma. Scale bar, 5 μm. (B) Time series of autophagosome biogenesis in the distal axon of hippocampal neurons grown 10 DIV. Arrowheads denote appearance of a GFP-LC3-positive punctum that grows into a ring structure. Scale bar, 1 μm. Also shown is the mean rate of autophagosome biogenesis in the mid versus distal axon (mean ± SEM; 5-6 DIV, n=12 neurons from 3 experiments; 10 DIV, n=8-9 neurons from 2 experiments; ns, not significant; **, p<0.01; two-way ANOVA with Tukey’s post-hoc test. (C) Time series of retrograde autophagosome motility in the axon of hippocampal neurons at 14 DIV. Arrowheads denote two retrograde autophagosomes. Scale bar, 5 μm. (D) Kymographs of autophagosome motility at various developmental stages. For clarity, the kymographs were traced and shown below each corresponding kymograph. Each autophagosome track is denoted by a different color. Horizontal bar, 5 μm. Vertical bar, 1 min. (E) Quantitation of autophagosome motility in the axon of hippocampal neurons (5-16 DIV; mean ± SEM; n=66 neurons from 4 experiments; ns, not significant; ***, p<0.001; one-way ANOVA with Tukey’s post-hoc test). (F) Quantitation of autophagosome flux over various developmental stages (mean ± SEM; 5-7 DIV, n=31 neurons from 3 experiments; 8-10 DIV, n=23 neurons from 3 experiments; 14-16 DIV, n=12 neurons from 3 experiments; ns, not significant; one-way ANOVA with Tukey’s post-hoc test). (G) Time series of bidirectional (left) and stationary (right) autophagosomes in dendrites. The retrograde direction is toward the right. Scale bar, 5 μm.

Similar to our previous observations in DRG neurons (Maday et al., 2012), autophagosomes that formed in the distal axon of hippocampal neurons initially move bidirectionally, then switch to robust processive, primarily unidirectional movement along the mid and proximal axon(Fig. 3C). The robust retrograde transport of autophagosomes along the axon was exhibited throughout development (5-16 DIV) with 87± 1.9% (± SEM) of axonal autophagosomes moving a net distance of ≥5 μm in the retrograde direction (Fig. 3D, E). Only~6% of axonal autophagosomes moved in the anterograde direction and ~7% exhibited non-processive bidirectional motility or remained stationary. Strikingly, autophagosome flux along the axon did not change with age of the culture (Fig. 3F).

We find that autophagosome dynamics in the axons of hippocampal and DRG neurons are remarkably similar. The mean autophagosome flux along the axon of hippocampal neurons (5-16 DIV) was similar to our previous reported valuein DRG neurons grown 2 DIV(Maday et al., 2012), 1.96± 0.14(± SEM) vs 1.76 ± 0.09 (± SEM) autophagosomes within 100 μm per min, respectively. Autophagosome motilityis robustly retrograde in both neuronal subtypes, 87 ± 1.9% (± SEM) in hippocampal neuronsas compared to82± 2.1% (± SEM)in DRG neurons; speeds are also very similar, with 0.62 ± 0.04 (± SEM) and 0.55 ± 0.06 (± SEM) μm/sec measured in hippocampal and DRG neurons, respectively. Importantly, constitutive biogenesis of autophagosomes is enriched in the distal axons of both hippocampal and DRG neurons. Thus, the overall paradigm of distal initiation followed by robust retrograde transport along the axon is similar between DRG and hippocampal neurons. The rates of biogenesis in the distal axon of hippocampal neurons, however, are ~4-fold lower as compared to DRG neurons. Since flux along the mid-axon is similar between these two neuronal types, autophagosome biogenesis may be less tightly restricted to the distal axon in synaptically-connected neurons as compared to developing axons undergoing robust growth. There may be more formation along the mid-axon of hippocampal neurons than is detected in our assay due to low formation rates distributed over the cumulative distance of the axon. Consistent with this, we observed lower autophagosome densities in the distal region of hippocampal neurons as compared to previous observations in DRG neurons (Maday et al., 2012).

We also observed autophagosomes in the dendrites of hippocampal neurons. In contrast to theprocessive retrograde motility in axons, autophagosomes in dendritesexhibited predominantly nonprocessive bidirectional and stationary transport (Fig. 3G). Of the 60 autophagosomes that could be unambiguously trackedwithin dendrites from 29 neurons, 61.7% were stationary, 16.7% exhibited bidirectional motility, 15.0% movedprocessively in the anterograde direction, and only 6.7% moved processively in the retrograde direction. Thus, the motility of autophagosomes within dendrites is strikingly different from what weobserved along the axon and may be due to the mixed microtubule polarity within dendrites (Baas et al., 1988).

The endoplasmic reticulum is the major membrane source for neuronal autophagosomes

Multiple organelles have been implicated as the source of membrane for autophagosomes, including plasma membrane(Hollenbeck, 1993; Ravikumar et al., 2010),Golgi(van der Vaart and Reggiori, 2010), mitochondria(Hailey et al., 2010), and endoplasmic reticulum (ER)(Hamasaki et al., 2013; Hayashi-Nishino et al., 2009; Yla-Anttila et al., 2009). For neurons, generating an autophagosome in the distal axon poses a unique challengedue to the confined nature of this space that is largely devoid of the Golgi. We performed dual-color imaging of Atg13 along with various membrane markers to determine the origin of the autophagosome membrane in primary DRG neurons. We utilized two markers, the src kinase Lyn and CellMaskOrange, to label the plasma membrane as well as plasma membrane-derived internal compartments (Fig. 4A, B). Atg13 puncta appearing in the distal axonwere negative for both plasma membrane markers (Fig. 4A, B). Thus, plasma membrane-derived material is not a likely source for neuronal autophagosome membrane in the distal axon under basal conditions.

Figure 4. Plasma and mitochondrial membrane are not incorporated into nascent autophagosomes.

(A) Time series of mCherry-Atg13 and plasma membrane marker Lyn-GFP in the distal axon of a DRG neuron. Atg13 is pseudocolored green for consistency throughout the figure. The retrograde direction is toward the right. (B) Time series of GFP-Atg13 and plasma membrane dye CellMask Orange. The retrograde direction is toward the right. (C) Time series of GFP-Atg13 and mitochondrial marker DsRed2-mito. The retrograde direction is toward the left. Throughout the figure, green arrowheads mark autophagosome biogenesis events. Dashed circles designate the location of the biogenesis event to demonstrate the absence of plasma and mitochondrial markers. Despite the absence on Atg13-positive puncta, plasma membrane markers are capable of labelinginternal organelles (open red arrowheads). Scale bars, 1 μm.

To address the possibility that mitochondria may supply membrane to nascent autophagosomes, we performed dual-color imaging with GFP-Atg13 and DsRed2-mito. We did not observe significant colocalization between Atg13 and mitochondria (Fig. 4C). While there was occasional transient overlap, there was no stable association between these compartments as assessed by kymographs (Fig. 5B). Thus, mitochondria are not a primary source of membrane for neuronal autophagosomes under basal conditions.

Figure 5. Neuronal autophagosomes form on or near the ER in the distal axon.

(A) Time series of mCherry-Atg13 and ER marker GFP-Sec61β in the distal axon of a DRG neuron. Yellow arrowheads denote autophagosome biogenesis events that colocalize and co-migrate with the ER. Green arrowheads denote regions of ER that colocalize with Atg13. The retrograde direction is toward the right. Linescans drawn across the distal axon (position marked with dashed lines) follow a single biogenesis event and indicate overlappingpeaks of intensity between Atg13 and the ER. Scale bar, 1 μm. See also Movie S2. (B) Corresponding kymographs betweenAtg13 and plasma membrane, mitochondria or ER. Atg13 co-migrates with the ER, suggesting that these compartments are associated. Since not all biogenesis events were captured at the start of Atg13 signal, the duration of the biogenesis event may vary. Horizontal bar, 1 μm. Vertical bar, 1 min. (C) Scatter plot of the Pearson’s Correlation Coefficient measured from corresponding kymographs of Atg13 and plasma membrane, mitochondria or ER (mean± SEM;plasma membrane, n=7 biogenesis events from 4 neurons from a representative experiment; mitochondria, n=11 biogenesis events from 7 neurons from a representative experiment; ER, n=20 biogenesis events from 11 neurons from 3 experiments; ns, not significant;*,p<0.05; **,p<0.01; one-way ANOVA with Tukey’s post-hoc test). Each point on the scatter plot represents one biogenesis event. Note that the sequential capture of images used here limits the maximum value of the Pearson’s Correlation Coefficient, even for colocalized signals.

In contrast, we did observe colocalization between Atg13 and the ER translocon subunit Sec61β. In the distal axon, Atg13 puncta appeared on or near ER structures labeled with Sec61β (Fig. 5A and Movie S2). Line scans across the distal tip consistently showed peaks of Atg13 coincident with some but not all peaks of Sec61β suggesting that specific subdomains of the ER are primed for autophagosome formation (Fig. 5A). Further, Atg13 and Sec61βoften co-migrated in the distal axon (Fig. 5A), providing compelling evidence that these structures colocalize and are dynamically connected. Interestingly, as the Atg13 punctum changed shape, the underlying ER alsochanged shape to match that of the Atg13 punctum. To further demonstrate coordinated motility between Atg13 and the ER, we generated kymographs from each biogenesis event. While there was no overlap between Atg13 and the plasma membrane or mitochondria, Atg13 and Sec61β colocalized and moved together in a coordinated fashion (Fig. 5B). Together, our results demonstrate an association between the ER and developing autophagosomes.

To quantitate the degree of colocalization between Atg13 and the ER, we measured the Pearson’s Correlation Coefficient between corresponding kymographs of Atg13 and Sec61β. The mean Pearson’s Coefficient value between Atg13 and the plasma membrane or mitochondria was negative and nearly zero, respectively (Fig. 5C). However, the mean value measured between Atg13 and Sec61β was significantly positive indicating colocalization between developing autophagosomes and the ER (Fig. 5C). As a control, we inverted the ER kymograph horizontally and observed a significant decrease in the mean Pearson’s Coefficient value (Fig. S3). Thus, colocalization between Atg13 and Sec61β is not simply due to the fact that the ER is widely distributed within the distal axon tip; rather there is a specific association between these two compartments. Together, these data support a key role for the ER in providing membrane to developing autophagosomes in primary neurons under basal conditions.

While the majority of biogenesis events originate from the ER, we noticed a minor population of events that undergo non-conventional forms of biogenesis, appearing to arise from pre-existing autophagosome rings. We occasionally observed “buds” appearingfrom pre-existing rings thatremain attached to the parental ring, moving together in the distal axon (Fig. S4). We also noticed the presence of Atg5 puncta associated withLC3 ring structures (Fig. S4). These results suggest that autophagosome rings may sometimes nucleate other smaller autophagic structures.

Neuronal autophagosomes form at DFCP1-positive subdomains of the ER

We next aimed to determine which subdomain of the ER plays a role in autophagosome formation. First, we explored the possibility that ER exit sites (ERES), a region rich in COPII-mediated membrane budding events, contributed membrane to nascent autophagosomes. Upon expression of GFP-Sec16L, a COPII coat assembly protein(Bhattacharyya and Glick, 2007), we noticed a striking gradient of ERES along the axon of DRG neurons. Sec16L puncta were concentrated in the cell soma and proximal axon, and decreased in density in the mid-axon and distal tip (Fig. 6A). Kymographs of Sec16L mobility over time showed that ERES moved bidirectionally within a confined region (within 2-3 μm) and most did not traverse large distances (>5 μm) within the axon (Fig. 6A). Kymographs also clearly displayed the decreasing gradient of ERES from the cell soma to the distal tip, with few ERES in the axon terminal. The same gradient was observed with another ERES marker, Sec24D (data not shown). These observed gradients along the axon are consistent with the cell soma being the primary site of protein synthesis in primary neurons.

Figure 6. Neuronal autophagosomes do not form at ER exit sites.

(A) Still images and corresponding kymographs of GFP-Sec16L in primary DRG neurons show a decreasing gradient of ERES from the cell soma to the distal tip of the axon. ERES are enriched in the cell soma and proximal axon, and depleted from the distal axon. Kymographs show bidirectional/stationary motility of ERES along the axon and clearly display the decreasing density of ERES from cell soma to tip. Horizontal bar, 1 μm. Vertical bar, 1 min. (B) Time series of mCherry-Atg13 and GFP-Sec16L in the distal axon. Red arrowheads denote biogenesis events that form at sites distinct from or absent of ERES (green arrowheads). Scale bar, 1 μm. (C, D) Kymographs generated from the boxed regions in B. In C, the motility of Atg13 is not correlated with that of Sec16L, indicating that autophagosome biogenesis occurs independent of ERES. In D, autophagosomes are formed in the absence of ERES. Horizontal bar, 1 μm. Vertical bar, 1 min.

The limited presence of ERES in the distal axon suggests that autophagosomes may not arise from ERES, which we confirmed by dual color live-cell imaging of Atg13 and Sec16L (Fig. 6B-D). In the event that an autophagosome was formed near an ERES, kymograph analysis showed that their motility was neither correlated nor overlapping (Fig. 6B, C). Thus, the ERES and nascent autophagosomes are distinct structures and autophagosomes are not generated from ERES in primary neurons.

Next we asked whether autophagosomesformed at DFCP1-positive subdomains of the ER. DFCP1 is an ER-localized PI(3)P-binding protein that in mammalian cells is thought to create a platform structure (the omegasome) from which the isolation membrane will nucleate from the ER(Axe et al., 2008). We imaged only those neurons expressing low levels of DFCP1. Consistent with previous reports(Axe et al., 2008), in DRG neurons DFCP1 initially appears as a punctum that grows into a ring seen as a flattened disc as it rotates (Fig. 7A). The ring structure created with DFCP1 is a true ring unlike that of the autophagosome, whichis a closed spherical organellethat only appearsas a ring due to confocal optical sectioning of this large organelle. The DFCP1 ring then collapses rather dramatically with a lifetime of ~6 min (Fig. 7A). The progressive increase in diameter and intensity followed by collapse can also clearly be seen with kymograph analysis (Fig. 7A).

Figure 7. Neuronal autophagosomes form at DFCP1-positive subdomains of the ER.

(A) Time series and corresponding kymograph of the assembly and disassembly of GFP-DFCP1-positive structures in the distal axon of a DRG neuron. The retrograde direction is toward the right. Horizontal bar, 1 μm. Vertical bar, 1 min. (B)Time series of mCherry-Atg13 and GFP-DFCP1 in the distal axon. Red and green arrowheads denote the ordered assembly of Atg13 followed by DFCP1 and ordered disassembly of Atg13 followed by DFCP1. The duration of overlap between these two markers is denoted with yellow arrowheads. The retrograde direction is toward the right. Scale bar, 1 μm. See also Movie S3. (C) Kymographs generated from the movie in B show colocalization and comigration between Atg13 and DFCP1. Horizontal bar, 1 μm. Vertical bar, 1 min. (D) Paired plot of DFCP1 biogenesis rates in the cell body, axon and distal tip from the same neuronshowing an increasing gradient of DFCP1 structures from cell soma to distal tip (n=6 neurons from 2 separate experiments). Movies were obtained for 10 min in each region. Only neurons with low DFCP1 expression were measured as overexpression of DFCP1 lead to uncoupling of the gradient. (E) Model for autophagosome formation in primary neurons.

We found that neuronal autophagosomes originate from DFCP1-labelled regions of the ER. We observed an ordered recruitment of these factors in the assembly of the nascent autophagosome. Atg13 appearance was followed by recruitment of DFCP1, and Atg13 disassembled prior to collapse of the DFCP1 ring (Fig. 7B and Movie S3). In contrast to the precise colocalization observed between Atg13 and LC3 puncta, Atg13 puncta appeared either inside or beside the DFCP1 ring (Fig. 7B and Movie S3). Kymograph analysis showedco-migration between Atg13 and DFCP1, indicating that these structures are associated during autophagosome formation (Fig. 7C). Thus, autophagosome biogenesis in primary neurons utilizes DFCP1-positive subdomains of the ER onto which assembly factors are recruited and disassembled in an ordered fashion.

Since we observed an enrichment of autophagosome biogenesis events in the distal axon (Fig. 2), we determined the distribution of DFCP1-positive biogenesis events within the neuron. Movies were obtained from the cell body, mid-axon and distal axon tip of the same neuron and the number of DFCP1 biogenesis events in each domain was quantitated. DFCP1 biogenesis events were enriched in the distal tip of the axon, while DFCP1 biogenesis events in the cell soma and axon were infrequent (Fig. 7D). Thus, consistent with the spatially regulated autophagosome biogenesis we observed (Fig. 2), the distribution of DFCP1-positive ER structures is also enriched in the distal axon.

Discussion

Using live-cell imaging, weinvestigated the spatiotemporal dynamics of autophagosome assembly in primary neurons in vitrounder basal growth conditions. We find that autophagosome formation in primary neurons involves a highly ordered pathway that proceeds with stereotypical kinetics. In DRG neurons, new autophagosomes are continuouslygenerated at a rate of ~0.6per min; formation of a single autophagosometakes4-6 min. Hippocampal neurons also exhibit robust biogenesis of constitutive autophagosomes albeit at a ~4-fold slower rate. We find that neuronal autophagosomes originate from specialized subdomains of the ER marked by the formation of DFCP1-positive ring structures. Remarkably, autophagosome biogenesis in primary neurons is spatially regulated along the axon. While autophagosomes can form in the cell soma and mid-axon, >80% of biogenesis occurs at the distal axon tip. This spatial specificity echoes the DFCP1 gradient observed along the axon. Importantly, the spatial regulation observed in DRG neurons was also observed in hippocampal neurons, including cultures establishing synaptic connections.

Consistent with previous studies in nonpolarized cells(Itakura and Mizushima, 2010; Koyama-Honda et al., 2013; Suzuki et al., 2001; Suzuki et al., 2007), we observed an ordered recruitment of Atg13 and Atg5 followed by LC3 onto nascent autophagosomes (Fig. 7E). We note that Atg13 and Atg5 recruitment occurs almost simultaneously within the time resolution of our experiment (1 frame every 2 s) (Fig. 7E). Atg13 puncta appear prior to DFCP1 and by extension, Atg5 is recruited prior to DFCP1. Appearance of Atg5 simultaneously with Atg13 and upstream of DFCP1 is unexpected considering genetic studies have functionally ordered Atg5 downstream of both Atg13 and DFCP1(Itakura and Mizushima, 2010). These results suggest thateither these proteins localize to the autophagosome assembly site prior theirfunction or that thebiochemical reaction ratesexceed our imaging resolution. While Atg13 and Atg5 arrive at the nascent autophagosome at the same time, Atg5 disassembles from the nascent autophagosome prior to loss of Atg13 (Fig. 7E). DFCP1 exit follows Atg13 decay.

These events proceed with stereotypical kinetics. Atg13 and Atg5 have an average lifetime of ~3.3 and ~3.0 min, respectively, on the nascent autophagosome, peaking ~1.3 and ~1.0 min, respectively, after initial appearance. LC3 initially undergoes a lag phase, followed by a robust rise to maximum intensity within 1.5 min. By the time LC3 reaches peak intensity, Atg5 has already begun a sharp decay. Thus, the formation of autophagosomes in primary neurons involves a highly ordered assembly and disassembly of proteins at the nucleation site.

The source of autophagosome membrane has been the subject of contentious debate. Autophagosomes could form de novo from newly synthesized lipids or they could arise from pre-existing organelles. We find that nascent autophagosomes are associated with the ER in the distal axon as evidenced by Atg13 colocalization and co-migration with the ER marker Sec61β. Further, Atg13 puncta overlapped with DFCP1-positive subdomains of the ER. Our results support a model in which the ER supplies membrane to developing autophagosomes under basal conditions in primary neurons (Fig. 7E). Our data are in agreement with both EM tomography (Hayashi-Nishino et al., 2009; Yla-Anttila et al., 2009) and colocalization studies(Axe et al., 2008; Itakura and Mizushima, 2010) implicating ER as the primary source for autophagosome membrane in other cell types. Activated Ulk1 complex is thought to be recruited to the ER along with the PI(3)P kinase complex, leading to local production of PI(3)P(Itakura and Mizushima, 2010; Matsunaga et al., 2010; Mizushima et al., 2011; Weidberg et al., 2011). DFCP1 is then recruited to P1(3)P-enriched domains and organizes a platform structure (the omegasome) at the ER from which the isolation membrane will be nucleated(Axe et al., 2008). 3D EM tomography provides strong evidence that the ER forms a cradle around and is connected to the isolation membrane during autophagosome formation(Hayashi-Nishino et al., 2009; Yla-Anttila et al., 2009).

By contrast, other reports have implicated the plasma membrane(Hollenbeck, 1993; Ravikumar et al., 2010), mitochondria(Hailey et al., 2010), ER-mitochondrial contact sites(Hamasaki et al., 2013), ERES(Graef et al., 2013), ERGIC(Ge et al., 2013), Golgi(van der Vaart and Reggiori, 2010)and recycling endosomes(Longatti et al., 2012; Puri et al., 2013) in autophagosome formation. Many of these studies were performed under stress-induced nutrient deprivation, suggesting there may be differential membrane sources for constitutive versus stress-induced autophagy. During metabolic stress, autophagy may be less discriminating in its membrane derivation and source lipid from various organelles.

Membrane origin may also be cell-type specific. In a yeast cell 5 microns wide, the Golgi may provide membrane for autophagosomes since they form in close proximity. However, in primary neurons, the spatial landscape is dramatically different. The Golgi is concentrated in the cell soma and thus unlikely to provide membrane for autophagosomes generated at the distal axon,at distances that can reach up to 1000 microns away.

Our most striking observation is the spatial regulation of autophagosome biogenesis in primary neurons. While autophagosomes were present throughout the neuron, along the axon, in the cell body and dendrites [and see(Bunge, 1973; Hernandez et al., 2012; Lee et al., 2011; Maday et al., 2012; Yue, 2007)], autophagosomeformation preferentially occurred in the distal tip of the axon; few were generated along the mid-axon or in the cell soma (Fig. 7E). The distal enrichment of autophagosome biogenesis observed in developing DRG neurons was paralleled in synaptically-connected hippocampal neurons, albeit the rates of distal formation in hippocampal neurons were lower than those observed in actively growing DRG neurons. Since flux along the mid-axon was remarkably similar between these neuronal subtypes, the difference in biogenesis rates suggests that autophagosome formation in synaptically-connected hippocampal neurons may not be as tightly restricted to the distal axon as in developing DRG neurons. While there may be higher levels of formation of autophagosomes along the mid-axon of hippocamal neurons, low biogenesis rates distributed over the cumulative distance of the axon makes observing these events rare.

This spatial regulation indicates a high degree of compartmentalization within the neuron. While ERES are concentrated in the cell soma and few in the axon, DFCP1 and autophagosome biogenesis is enriched distally (Fig. 7E). Thus, there are opposing gradients along the axon, emphasizing the striking ability of neurons to compartmentalize functions. While local translation has been observed in the distal axon(Holt and Schuman, 2013), the bulk of protein synthesis occurs in the cell soma. However,>80% of autophagosome formation occurs distally. The enrichment of DFCP1-positive structures in the distal tip suggests that perhaps the ER in the distal axon is more specialized and primed for autophagosome production. Alternatively, autophagy regulators such as mTOR may also share a polarized distribution or activity along the axon, accounting for this distal enrichment of autophagosome formation. Compartmentalization within the neuron is further evident from our observations that autophagosome dynamics in dendrites are distinct from those in axons. These differences may result from the underlying mixed polarity of dendritic microtubules as compared to the unipolar organization of axonal microtubules, or may reflect differential regulation; future work will be needed to explore these possibilities.

Our results raise an interesting question as to why autophagosomes are preferentially generated at the distal end of the axon. The axon terminal is a region rich in activity as it is actively remodeledduring phases of extension and retraction during neurite outgrowth. As a result, there is an increased demand for membrane recycling and organelles in this region might be more susceptible to damage. EM studies have shown autophagosomes in various stages of developmentto be enriched in growth cones, particularly in regions undergoing retraction(Bunge, 1973). However, our observations inhippocampal neurons alsodemonstrate the distal enrichment of autophagosome formation,suggesting that this spatial regulation is not limited to DRG neurons actively extending processes. Thus, constitutive autophagy in the distal axon may act to counterbalance the anterograde flow of slow axonal transport. Cytoplasmic proteins synthesized in the cell soma reach the distal axon in a slow sustained stream that moves 1-10 mm per day(Brown, 2000; Scott et al., 2011). In a human motor neuron one meter in length, cytoplasmic proteins transported by slow transport would have agedfrom ~100 days up to 2.7 yearsupon arrival in the distal tip. How these long-lived proteins are degraded or recycled once they reach the axon terminal is unclear. High autophagic activity in the distal axon may serve to counteract thedistal accumulation of aged proteinsover time.

Elevated levels of autophagy in the distal axon may also be required to recycleaged organelles that preferentially distribute to the distal axon. Mitochondria residing in the distal axon are older than those residing proximal to the cell soma (Ferree et al., 2013). Located further from primary sites of protein synthesis, organelles in the distal axon may be more susceptible to aging and damage since their proteins are less efficiently replenished. Consistent with this idea,higher levels of mtDNA damagehave been detected in distal axons in patients with HIV-related sensory neuropathy(Lehmann et al., 2011). Thus, elevated autophagy in the distal axon may provide a mechanism to recycle aged and damagedmitochondriain the distal region of long axons.

While autophagosome biogenesis canoccur in the mid-axon or cell soma, these events are infrequent under basal conditions. So how are damaged organelles and proteins along the mid-axon cleared?Perhaps the axon has a limited capacity for degradation. A striking feature that unifies many neurodegenerative diseases is the aberrant accumulation of protein aggregates along the axon that results in axonal retraction. The autophagic pathway is preferentially elevated in the distal axon in some disease models. For example, in a mouse model of excitotoxic neurodegeneration, autophagosomes preferentially accumulate in the distal axons of Purkinje neurons,at levels exceeding those in the cell soma or dendrites(Wang et al., 2006). A Purkinje cell-specific knockout of Atg7 is sufficient to cause degeneration of the axon terminal with little effect on dendrites,indicating a critical role for autophagy in the maintenance of axon terminals (Komatsu et al., 2007). Further, neocortical biopsies from Alzheimer’s diseasebrain exhibit a pronounced accumulation of autophagosomes in the axons as compared to the soma, with particular abundance in synaptic terminals(Nixon et al., 2005). Collectively, these data emphasize a role for the compartmentalized regulation of autophagybut suggest that beyond a certain threshold, the autophagic system is unable to effectively remove damaged proteins or dysfunctional organelles, rendering the neuron susceptible to disease.

Autophagy hasalso been linked to synaptic structure and function. Autophagy regulates presynaptic activity and vesicle release, NMJ structure, as well as receptor turnover in the postsynaptic membrane(Hernandez et al., 2012; Shehata et al., 2012; Shen and Ganetzky, 2009). In developing DRG neurons that do not form synapses, autophagosome formation was significantly enriched in the distal axon, perhaps due to the dynamic states of growth cone extension and retraction. In contrast, in synaptically-connected hippocampal neurons, constitutive autophagosome biogenesis is less tightly restricted to the distal axon, potentiallyoccurringat en passant synapses along the mid-axon as suggested by observations in dopaminergic neurons(Hernandez et al., 2012). Future work is required to address the possibility that stimulation of synaptic activity alters dynamics andincreases autophagosome formation along the mid-axon. Further, examination of autophagosome formation and dynamics in vivo will provide additional insights on the regulation of autophagy in developing versus mature systems undergoing physiological levels of synaptic activity.

Together, our results highlight the high degree of compartmentalization and spatial regulation imposed on cellular pathways within the neuron. We find that autophagosome biogenesis in primary neurons is a polarized process that is spatiotemporally regulated along the axon. Our findings here provide insights into the unique vulnerability of neurons to increased protein aggregation along the axon, resulting in neuronal dysfunction and degeneration.

Experimental Procedures

Reagents

GFP-LC3 transgenic mice, strain name B6.Cg-Tg(CAG-EGFP/LC3)53Nmi/NmiRbrc(Mizushima et al., 2004), were obtained from the RIKEN BioResource Center in Japan.Constructs and antibodies are detailed in the supplemental experimental procedures.

Imaging of primary neurons

Primary neuron culture

Dorsal root ganglia were cultured as previously described by Maday et al. (2012); see supplemental experimental procedures for details. For hippocampal neurons, the hippocampus was dissected from E15.5 mice of either sex, dissociated in 0.25% trypsin, and triturated through a Pasteur pipet, as described in the supplemental experimental procedures. To determine the directionality of single axons during live-cell imaging, GFP-LC3-positive neurons were diluted 1:10 with GFP-LC3-negative neuronsisolated from non-transgenic littermates. With this dilution, single GFP-LC3-positive neurons could be observed and the directionality of the axon was determinedbased on criteria established by Kaech and Banker(Kaech and Banker, 2006). All animal protocols were approved by the Institutional Animal Care and Use Committee at the University of Pennsylvania.

Live-cell imaging

Live-cell imaging was performed as described in Maday et al. (2012); see the supplemental experimental procedures. Biogenesis movies in the mid and distal axon were obtained whenever possible from the same neuron. For hippocampal cultures, axons were selected for imagingbased on morphologic criteria(Kaech and Banker, 2006): thin and uniform caliber with no swellings or varicosities along the length. Dendrites were also identified based on morphologic criteria(Kaech and Banker, 2006): 200-300 μm lengths that tapered with distance from the cell soma. Kymographs were generated using the MultipleKymograph plugin in ImageJ FIJI using a line width of 3 or 5.

Image Analysis

Quantitation of biogenesis along the axon

Quantitation of intensity profiles is described in the supplemental experimental procedures. Movies were acquiredwithin ~50 μm (for DRG neurons) or ~70 μm (for hippocampal neurons)of the distal end of the axon, defined here as the distal tip, over an ~80 μm window along the mid-axon that spanned 500-1000 μm in length(length for DRG neurons), or in the cell soma,~25 μm in diameter. For all DFCP1 experiments, a single distal tip, mid-axon and cell soma movie was obtained from the same neuron. For the Atg13 and Atg5 experiments, some but not all movies were obtained from the same neuron. Only neurons with low Atg13 or Atg5 expression were imaged in order to observe specific events of biogenesis. Only neurons with a mean intensity of ≤6.2 arbitrary units (corrected for background) for cytosolic DFCP1 intensity were quantified. We imaged neurons with flat axon tips in order to accurately count the total number of biogenesis events in the region. Biogenesis events were defined by changes in intensity and size of Atg5, Atg13, LC3 or DFCP1over time as measured in Figure 1 and Figure 7A. Any biogenesis event within each domain was counted and normalized for movie duration.

Quantitation of axonal transport in hippocampal neurons

Using FIJI, kymographs (line width 3) were generated from movies along the axon >50 μm from the cell soma and >100 μm from the distal tip. For each kymograph the percentage of autophagosomes moving in the net retrograde direction (displacement of ≥5 μm within the 5 min imaging window) versus net anterograde direction (displacement of ≥5 μm within 5 min) was determined. Vesicles that did not move a net 5 μm within the 5 min imaging window were classified as bidirectional or stationary. Flux (number of autophagosomes within 100 μm per min) was determined from each kymograph as the sum of retrograde, anterograde and bidirectional/stationary vesicles normalized by kymograph length and time. Average velocity of retrograde autophagosomes was calculated as displacement divided by total time taken.

Quantitation of dendritic transport in hippocampal neurons

A comprehensive analysis of dendritic autophagosome motility was limited due to the increased depth of the dendrite combined with brighter microtubule labeling byGFP-LC3. Thus, only autophagosomes that could be tracked unambiguously were categorized as retrograde (processive movement in the retrograde direction; displacement of ≥5 μm), anterograde (processive movement in the anterograde direction;displacement of ≥5 μm), stationary (displacement of <5 μm), or bidirectional (processive movement of 5 μm in both anterograde and retrograde directions).

All image measurements were obtained from the raw data. Procedures describing immunostain, line scans and Pearson’s Correlation Coefficient measurements are provided in the supplemental experimental procedures. GraphPad Prism was used to plot graphs and perform statistical tests; statistical tests are denoted within each figure legend. Images were prepared in FIJI; contrast and brightness was adjusted equally to all images within a series. Figures were assembled in Adobe Illustrator.

Highlights.

• Autophagosomes in primary neurons form via an ordered assembly pathway

• Neuronal autophagosomes form at DFCP1-positive subdomains of the ER

• Autophagosomebiogenesis in primary neurons is enriched in the distal axon

• Constitutive autophagy follows a compartmentalized pathway in primary neurons

Abbreviations List

DRG

dorsal root ganglion

LC3

microtubule-associated protein light chain 3

Atg

autophagy-related

ERES

ER exit sites

DFCP1

double FYVE domain-containing protein 1

PI(3)P

phosphatidylinositol 3-phosphate