|Home | About | Journals | Submit | Contact Us | Français|
Macroautophagy, a major pathway for organelle and protein turnover, has been implicated in the neurodegeneration of Alzheimer's disease (AD). The basis for the profuse accumulation of autophagic vacuoles (AVs) in affected neurons of the AD brain, however, is unknown. In this study, we show that constitutive macroautophagy in primary cortical neurons is highly efficient, as newly formed autophagosomes are rapidly cleared by fusion with lysosomes, accounting for their scarcity in the healthy brain. Even after macroautophagy is strongly induced by suppressing mTOR kinase activity with rapamycin or nutrient deprivation, active cathepsin-positive autolysosomes rather than LC3-II-positive autophagosomes predominate, implying efficient autophagosome clearance in healthy neurons. By contrast, selectively impeding late steps in macroautophagy by inhibiting cathepsin-mediated proteolysis within autophagosomes with cysteine- and aspartyl-protease inhibitors caused a marked accumulation of electron-dense double membrane-limited AVs, containing cathepsin D and incompletely degraded LC3-II in perikarya and neurites. Similar structures accumulated in large numbers when fusion of autophagosomes with lysosomes was slowed by disrupting their transport on microtubules with vinblastine. Finally, we find that the autophagic vacuoles accumulating after protease inhibition or prolonged vinblastine treatment strongly resembled AVs that collect in dystrophic neurites in the AD brain and in an AD mouse model. We conclude that macroautophagy is constitutively active and highly efficient in healthy neurons, and that the autophagic pathology observed in AD most likely arises from impaired clearance of AVs rather than strong autophagy induction alone. Therapeutic modulation of autophagy in AD may, therefore, require targeting late steps in the autophagic pathway.
The accumulation of specific cellular proteins within neurons is a pathologic hallmark of many major neurodegenerative diseases. Dysfunction within either of two proteolytic pathways, involved in clearing abnormal or obsolete cellular proteins, the ubiquitin-proteasome system (UPS) and the autophagic-lysosomal system, have been increasingly implicated in proteinopathies of Alzheimer's disease, Huntington's disease, Parkinson's disease and related disorders (Rubinsztein et al., 2005; Chu, 2006; Zhu et al., 2007). Macroautophagy, a major degradative pathway of the lysosomal system, plays a unique role in removing cellular organelles and protein aggregates that are too large to be degraded by the UPS (Rideout et al., 2004). In macroautophagy, hereafter referred to as autophagy, regional sequestration of cytoplasm within an enveloping double membrane structure, creates a vacuole, termed the autophagosome (De Duve and Wattiaux, 1966; Klionsky et al., 2008). Degradation of the cytoplasmic substrates is initiated when the autophagosome fuses with degradative compartments of the endosomal-lysosomal system (Berg et al., 1998). Constitutive autophagy in neurons prevents ubiquitinated proteins from accumulating and is essential for cell survival (Hara et al., 2006; Komatsu et al., 2006). In addition to exerting control over cellular protein quality, autophagy is induced during nutritional deprivation, trophic factor withdrawal, and other types of cell stress to protect cells against apoptosis by degrading non-essential cell constituents for energy (Boland and Nixon, 2006; Finn and Dice, 2006). Autophagy contributes to cell remodeling during differentiation but strong induction during development initiates a form of “autophagic” programmed cell death in some cell populations (Levine and Klionsky, 2004). Overactive or dysfunctional autophagy may also promote neuronal cell death in disease states (Nixon, 2006).
The scarcity of recognizable autophagosomes and other autophagic vacuole (AV) intermediates in healthy neurons has raised the possibility that baseline autophagy in neurons is normally low and that the frequent presence of AVs in neurodegenerative states reflects an induction of autophagy (Boellaard et al., 1989; Anglade et al., 1997; Kegel et al., 2000; Ko et al., 2005; Nixon et al., 2005). For example, in Alzheimer's disease, AVs accumulate in enormous numbers within grossly distended, “dystrophic” neurites of affected neurons and also appear in other parts of neurons. The relative contribution of heightened autophagy induction to the robust AV expansion in Alzheimer's disease is uncertain, as it is also possible that AVs might accumulate to high numbers when autophagosomal digestion is impaired. An equilibrium exists between autophagosome formation and clearance by lysosomes (Chu, 2006), termed autophagic flux, and it is now known that terminal stages of autophagosomal degradation require uncompromised vesicular trafficking, heterotypic organelle fusion and lysosomal function (Kovacs et al., 1982; Takeshige et al., 1992; Tanida et al., 2005; Kiselyov et al., 2007). Interestingly, a number of recent studies on various lysosomal storage disorders have found evidence of impaired AV clearance (Kiselyov et al., 2007; Settembre et al., 2008). Little is known about the responsiveness of neurons to modulation of the autophagic system. To begin to understand the underlying molecular basis for pathological autophagic patterns in neurodegenerative disease states, we investigated responses of primary cortical neurons to robust induction of autophagy. Alternatively, we interfered with clearance of autophagosomes by (i) impeding autophagosome trafficking and subsequent fusion with lysosomes or by (ii) inhibiting lysosomal proteolysis. Morphological and biochemical indices of autophagic states revealed that in healthy neurons autophagosomes are actively formed and efficiently cleared. Interference with AV clearance, rather than induction of autophagy alone, gives rise to patterns of autophagy pathology resembling those seen in Alzheimer's disease.
Primary cortical neurons were derived from Day 0 (0-12 hr postnatal) Sprague Dawley rat pups (Charles River, Wilmington, MA). Pup brains were harvested and placed in ice-cold Hibernate-E media (BrainBits LLC, Springfield, IL) where the meninges were removed and the cerebral cortices were dissected. Cortices were minced using a scalpel and dissociated by incubating the tissue in Hibernate-E media containing 15units/ml of papain (Worthington Biochemical's, Lakewood, NJ) for 30 mins at 37°C before triturating in Neurobasal medium (Invitrogen, Carlsbad, CA) containing 20% fetal bovine serum (Hyclone, Logan, UT) and DNAse (0.2μg/ml). Undissociated brain tissue was removed by passing the cell suspension through a 40μm cell strainer (Fisher Scientific, Suwanee, GA). Dissociated neurons were centrifuged at 200 × g for 3 min at RT and the pellet was resuspended in Neurobasal medium supplemented with B27 (2%), penicillin (100U/ml), streptomycin (100U/ml) and glutamine (0.5mM, all Invitrogen). Viable neurons were plated at a density of 100,000 cells per 13mm circular coverglass and 250,000 cells per well in 6-well tissue culture dishes, pre-coated with poly-d-lysine (50μg/ml, Sigma, St. Louis, MO), and incubated in a humidified atmosphere containing 5% CO2: 95% atmosphere at 37°C. Half of the plating media was replaced with fresh pen/strep-free media after three days. Serum-free, B27 supplemented Neurobasal medium ensured minimal growth of glial cells (<5%) after 5 days in culture.
After five days in vitro, cultured neurons were treated for 1 hr, 6 hr and 24 hr before fixation or harvesting. Treatment conditions and compounds used included rapamycin (10nM, L.C. Laboratories, Woburn, MA), vinblastine (10μM, Sigma), leupeptin (20μM, Sigma), pepstatin A (20μM, Sigma) and Earle's Balanced Salt Solution (EBSS; Invitrogen)
Cortical biopsy specimens were studied from a 74 year old male with a clinical diagnosis of dementia and prominent dyspraxia The patient underwent surgery for placement of a ventricular shunt and frontal cortex tissue obtained next to the shunt penetration site revealed numerous neurofibrillary tangles and cored amyloid plaques consistent with the neuropathologic diagnosis of Alzheimer's disease as previously described (Nixon et al., 2005). Tissue was fixed in 3% glutaraldehyde/0.1M phosphate buffer, pH 7.4, and postfixed in 1% osmium tetroxide in Sorensen's phosphate buffer. After dehydration in ethyl alcohol, the tissue was embedded in Epon (EMS, Fort Washington, PA). Tissue blocks were cut serially into ultrathin (0.06 μm) sections. Ultrathin sections were stained with uranyl acetate and lead citrate. Transgenic mice expressing the Swedish mutation of human APP (APPK670M/N671L) and mutant PS1 (PS1M140L; (Duff et al., 1996) were studied at 16 months of age. Ultrathin sections of the cerebral cortex were cut from Epon-embedded blocks and placed on copper grids for structural analysis.
Neurons were fixed for electron microscopy (EM) by removing culture medium, washing (x3) in 37°C supplement-free Neurobasal medium and adding 4% paraformaldehyde and 1% glutaraldehyde in 0.1M sodium cacodylate buffer (pH 7.2; Electron Microscopy Sciences (EMS), Hatfield, PA) for 3 hr at room temperature (RT). Following fixation, neurons were washed (x3) in cacodylate buffer, post-fixed in 1% osmium tetroxide and progressively dehydrated in a graded series of ethanol (50% - 100%) and embedded in Epon. Thin sections of 1μM were cut from the polymer, followed by ultrathin sections (70-80nm) using a Recheirt Ultracut S microtome and placed on copper grids for structural analysis. Grids were briefly stained with uranyl acetate and lead citrate before being examined with a Philips electron microscope (Model CM 10). Images were captured on a digital camera (Hamamatsu, model C4742-95) using Advantage CCD Camera System software (Advanced Microscopy Techniques Corporation, Danvers, MA).
Polymerase chain reaction (PCR)-amplified coding sequences of LC3 in GFP-LC3 (provided by Noboru Mizushima, Tokyo Medical and Dental University) was sub-cloned into a pDsRed-Monomer-C1 vector (Clontech, Mountain View, CA) to create DsRed-LC3, and verified by sequencing. The pEGFP-Endo reporter vector, which encodes a fusion protein containing the human RhoB GTPase, was obtained from BD Biosciences (Clontech Laboratories Inc., Mountain View, CA). Primary cortical neurons 3-4 days in vitro (DIV) plated in 35mm glass-bottom dishes were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's suggested conditions. Briefly, 2ml conditioned (pre-transfection) medium was replaced with transfection medium consisting of 1μg DNA, 5μl Lipofectamine 2000, 500μl Opti-Mem (Invitrogen), and 1.5 ml Neurobasal medium without B27. Neurons were incubated with transfection media for 30 min at 37°C, followed by replacement (x3) with fresh Neurobasal medium. Conditioned medium was re-added to the transfected neurons and maintained in the incubator for least 24 hr before treatments.
DsRed-LC3 transfected primary cortical neurons were incubated with 1μM BODIPY-pepstatin-FL (Invitrogen) in Neurobasal medium for 1 hr at 37°C followed by replacement with fresh Neurobasal medium (x2). Subsequently, Neurobasal medium was replaced with low fluorescence Hibernate medium (BrainBits) to reduce fluorescent background and cultures were placed in a 37°C humidified chamber with 5% CO2 on a Ziess LSM510 confocal microscope. Z-stack images were acquired using LSM 520 software.
Neurons (DIV 5) grown in 6-well plastic dishes (BD Bioscience, Franklin Lakes, NJ) were washed (x3) in phosphate buffered saline (PBS, pH 7.4) at RT, then scraped in 200μl/well of ice-cold lysis buffer (M-PER buffer containing Halt Protease Inhibitor Cocktail (1:100) and 50μM EDTA, Pierce, Rockford, IL). Protein concentration was determined using the Bradford Assay (Pierce) and samples were standardized to 1μg/μl using 70% trichloroacetic acid to precipitate cell lysates that were resuspended in equal volumes of lysis buffer. Sample loading buffer (2X) (62.5mM Tris-HCl (pH 6.8), 25% glycerol, 2% SDS, 0.01% bromophenol blue, 710μM β-mercaptoethanol) was added to cell lysates at a 1:1 ratio with lysis buffer before heating samples for 5 mins at 90°C. Tris-glycine gels (Invitrogen) were loaded with 25μg/well of protein and separated using a 10% gel for phospho-p70 S6 kinase (Thr 389), and total p70 S6 kinase (both from Cell Signaling) and a 16% gel to separate LC3-I and LC3-II. Separated proteins were transferred onto 0.2μm nitrocellulose membranes (Protran, Whatman, Florham Park, NJ) for 2 hr at 250mA (p70S6 kinase) or 8 hr at 100mA (LC3). Membranes were rinsed in Tris-buffered saline containing 0.1% Tween-20 (TBS-T) before being blocked at RT for 1 hr in a 5% non-fat milk/TBS-T solution. Primary antibodies were diluted in a 1% BSA/TBS-T solution at 1:250 for both p-p70 and total p70, and at 1:1000 for LC3 (primary antibody raised against rattus LC3 based on a GST-tagged recombinant protein (Yu et al., 2005). Membranes were incubated with primary antibody for 2 hr at RT, washed three times in TBS-T and then incubated for 1 hr at RT in secondary antibody solution (3% non-fat milk/TBST) containing alkaline phosphatase-conjugated secondary antibodies (Promega, Madison, WI) for p-p70S6 kinase and total p70S6 kinase blots, and a horseradish peroxidase-conjugated antibody for LC3 blots (GE Healthcare, Piscataway, NJ). Membranes were washed three times for 10 mins in PBS-Tween before chemiluminescent substrates (CDP-Star, Applied Biosystems, Forest City, CA; ECL, GE Healthcare) were applied and membranes were exposed to X-ray film. Densitometry of bands representing protein expression was done using Image-J (NIH Image v.1.63) Software from the NIH. For each immunoblot, the band intensity of each lane was normalized relative to the control lane. Subsequently, the % change between treatments was calculated based on the normalized values.
Neurons were fixed for immunocytochemistry analyses by removing culture medium, washing (x3) in PBS and adding 4% paraformaldehyde/PBS (pH 7.4) for 15 mins at RT. Neurons were permeabilized for 15 mins in 0.02% Triton-X/PBS and then blocked for 1 hr at RT with a 2% Fetal Bovine Serum/0.02% Triton-X/PBS blocking solution. Primary antibodies were made up in blocking solution and incubated for 2 hr with neuron-specific class III beta-tubulin (1:1000, TUJ-1, R & D Systems, Minneapolis, MN), overnight for LC3 (1:250, Nanotools, Teningen, Germany) and cathepsin D (1:5000, in house) at 4°C. Cells were washed (x3) in PBS for 10 mins prior to incubation with TRITC- and FITC-labeled secondary antibodies (1:1000/blocking solution, Invitrogen) for 1 hr at RT. Neurons were washed (x3) in PBS for 10 mins before mounting coverslips onto microscope slides with anti-fade Gelmount (Biomeda, Foster City, CA) and visualized using a Zeiss confocal microscope.
Neurons were fixed with 4% paraformaldehyde, 0.15% glutaraldehyde and 4% sucrose in 0.1M sodium cacodylate buffer in Neurobasal media for 30 mins at 37°C followed by incubation in fixative without Neurobasal at 4°C for 24 hr. Pre- and post-embedding immuno-EM were done in parallel, using the same primary antibody. For pre-embedding, cells were incubated with 0.1% sodium borohydride in PBS for 15 mins and washed (x4) in PBS for 10 mins. Cells were permeabilized with 0.05% Triton-X in PBS for 30 mins, washed in PBS (x4) for 10 mins, then blocked with 5% BSA and 5% Normal Goat Serum in PBS for 1 hr at 4°C. Cells were incubated with primary antibody polyclonal Cathepsin D (1:2000 in blocking buffer) at 4°C overnight followed by washing and incubation with ultra-small gold-conjugated goat-anti-rabbit secondary antibody (1:100 in blocking buffer) for 2 hr at RT. Ultra-small gold was silver-enhanced with Custom Ultra Small Kit (EMS) using the manufacturer's guidelines. Following silver-enhancement, cells were post-fixed with osmium tetroxide and processed for EM, as previously described (Yu et al, 2005). For post-embedding immuno-EM, cells were fixed with freshly made 3% paraformaldehyde containing 0.1% glutaraldehyde and 4% sucrose in 0.1M sodium cacodylate buffer. Following fixation, cells were washed, dehydrated, and embedded in Lowicryl K4M (Polysciences, Inc., Warrington, PA) and polymerized under UV light (360nm) at −35°C. Ultrathin sections were cut, mounted on Formvar and carbon coated nickel grids. After incubation with primary antibodies at 4°C overnight, gold conjugated secondary antibodies (18nm Colloidal Gold-AffiniPure Goat Anti-Rabbit IgG(H+L), Jackson ImmunoResearch Laboratories, Inc., West Grove, PA; 15nm Protein A Gold, Cell Microscopy Center, University Medical Center Utrecht, 35584 CX Utrecht, The Netherlands) were applied and stained with uranyl acetate and lead citrate by standard methods. Post-embedding immunogold electron microscopy on human brain was performed as previously described (Nixon et al., 2005).
The number of autophagic vacuoles per cell body were counted using EM images at direct magnification of 10,500x (64 μm2) for the various treatment conditions. Electron micrographs (25-100 images per treatment condition) were examined and values are expressed as AVs per field. For our analysis, we identified AVs as compartments of 0.5μM in diameter or larger. Autophagic vacuoles were further classified according to whether they were autophagosomes or autolysosomes based on two qualities: autophagosomes have double membranes and uncompacted cytoplasmic material including organelles such as mitochondria and ribosomes, whereas autolysosomes are single or double-membrane-limited vesicles with densely compacted amorphous or multilamellar contents. All numerical values are expressed as mean +/− SEM.
As a baseline for studies of autophagy induction, we used neurons from postnatal rat pups (P0) grown for five days in a nutrient-rich condition in the presence of serum supplement, B27. Exposure of these cells to rapamycin (10nM) for 1 hr rapidly induced autophagy, as evidenced by a 78 ± 4.5% decrease in mTOR-mediated phosphorylation of p70 S6 kinase (ratio of phosphorylated p70 to total p70) as compared to untreated neurons (Fig. 1A, B). Decreased mTOR activity was sustained during rapamycin treatment for 6-24 hr (Fig. 1A, B). Post-translational modification of cytosolic LC3-I (Light Chain 3 of Microtubule Associated Protein 1A/B) to LC3-II is followed by LC3-II translocation to nascent autophagosomal membranes, which is a useful index of autophagosome formation (Mizushima, 2004). Rapamycin elevated LC3-II levels by 55 ± 3% at 1 hr and by 80-90% from 6-24 hr (Fig. 1D). Tubulin levels, serving as an additional independent control protein, did not change significantly under any of the conditions of analysis in this study.
Cortical neurons cultured in nutrient-rich media for five days rarely contained autophagosomes identified by LC3 labeling or by ultrastructural analysis (Fig. 1E, Fig. 2A, B). Rapamycin (10nM) treatment of neurons for 24hr caused LC3 immunoreactivity in perikarya to redistribute from the cytosol to vesicular structures (Fig. 1F). To visualize the maturation of autophagic vacuoles, we transfected neurons with DsRed tagged LC3 (DsRed-LC3) and loaded them with BODIPY-pepstatin-FL-a fluorescent probe that binds to active cathepsin D, and is an indicator of protease activity and indirectly of vesicle acidification. LC3-positive vesicular compartments were relatively uncommon while BODIPY-positive lysosomes were abundant in neurons under basal conditions (Fig. 1 G). In rapamycin-treated neurons, DsRed-LC3 redistributed from the cytosol to AVs (Fig. 1H). The stronger vesicular LC3 in DsRed-transfected neurons compared to endogenous LC3 detected by immunostaining (compare Figure 1 F and H) can be explained by transient overexpression of DsRed-LC3 and its relative stability in the lumen of the AV. Despite strong autophagy induction for 6h with rapamycin (10nM), the vast majority of the DsRed-LC3-positive compartments contained abundant active cathepsin D, indicating that fusion with lysosomes occurs rapidly after autophagosome formation (Fig. 1I). These AVs contained LC3 on the limiting membrane as well as incompletely degraded DsRed-LC3 in the lumen (Fig. 1I, left inset), and were classified as autophagolysosomes, or autophagosomes that have undergone fusion with lysosomes and hereafter referred to as autolysosomes. Under the same conditions, AVs that predominantly contained cathepsin but displayed only traces of undegraded ds-Red LC3 in the lumen were also observed - representing a more mature autophagosome or autolysosome (Fig. 1G, I). EM analyses of rapamycin-treated neurons revealed markedly increased proportions of large (0.5 to 1μM) relatively clear vacuoles in cell bodies containing scattered amorphous material after 1hr (Fig. 2C) and through to 24hr (Fig. 2E). Overall, the ultrastructure of most AVs in rapamycin-treated neurons resembled lysosomes observed in untreated neurons (Fig. 2G) that were enlarged. We collected counts of autophagosomes and autolysosomes on a series of electron micrograph images of cell bodies from rapamycin-treated and control neurons and found that autolysosomes were significantly greater in number (8.21 ± 1.33 per field, p<0.001) than autophagosomes (2.42 ± 0.56 per field, p<0.001) compared to controls (0.11 ± 0.15 per field). Furthermore, in immuno-EM analyses, almost all AVs accumulating after rapamycin treatment contained immunogold-labeled cathepsin D, many at levels similar to those in lysosomes of neurons under basal conditions (Fig. 2H). These results indicate that neurons initiate and sustain autophagy induction when mTOR is inhibited; however, they do not accumulate large numbers of autophagosomes, which are efficiently converted to autolysosomes by lysosomal hydrolases (Tanida et al., 2004).
Lysosomal proteolysis is predominantly mediated by cysteine and aspartyl proteases, which are inhibited by leupeptin and pepstatin, respectively. In neurons treated with leupeptin, pepstatin, or both, the levels of total p70S6k and p-p70S6k (phospho-Thr 389) remained comparable to those in untreated neurons (Fig. 3A, B), indicating that cathepsin inhibition does not lead to autophagy induction through mTOR inhibition. However, even in the absence of mTOR inactivation, leupeptin (20μM) increased LC3-II levels more than 4-fold over that of control LC3-II levels by 24 hr (Fig. 3C, D). LC3-positive compartments also became significantly more numerous after co-treatment with leupeptin and pepstatin (Fig. 3E), considerably exceeding the numbers seen in rapamycin-treated cultures (Fig. 1F). The cysteine protease inhibitor, leupeptin, did not alter cathepsin D activity in DsRed-LC3 vesicles (Fig. 3E), but DsRed-LC3 levels were higher in most vesicles compared to those after rapamycin treatment (compare Fig. 1 H, I and Fig. 3E). The pEGFP-Endo reporter vector encodes a fusion protein containing the human RhoB GTPase, and labels endocytic compartments. To confirm delayed degradation of endosomes as well as autophagosmes when lysosomal proteases are impaired, we transfected primary neurons with GFP-Endo and immunostained for LC3 and cathepsin D. GFP-Endo does not colocalize with cathepsin D under normal conditions (3F, left panel), but occasionally colocalizes with LC3-positive AVs - although this was rare since vesicular LC3 is not often observed in untreated neurons (3F, middle panel). After treatment with leupeptin and pepstatin, vesicles were identified that were both GFP-Endo-positive and LC3-positive (Figure 3G middle panel, arrows). Most of these profiles contained cathepsin D (right panel, inset, arrow) similar to the DsRed-LC3/BODIPY-pepstatin-positive AVs that accumulated under the same condition. The presence of endosomal and autophagosomal marker proteins with cathepsin D, indicates many of these compartments are amphisomes (Gordon and Seglen, 1988). Also, GFP-Endo-positive vesicles containing cathepsin D (Figure 3G left panel, inset, arrows) were detected, which confirmed impaired degradation of endocytic as well as autophagic compartments, when lysosomal proteases are inhibited. These results indicate that cathepsin inhibition promotes accumulation of autophagic compartments including amphisomes, and that degradation within these compartments is impaired. Numbers of predominantly double-membrane-limited AVs containing amorphous electron-dense material began to rise by 6 hr in the cell bodies (Fig. 4A) and some neurites (Fig. 4B) of leupeptin-treated neurons and were very abundant in both perikarya and neurites by 24 hr of treatment with leupeptin (Fig. 4C, D; 5.85 ± 1.46 per field, p <0.01 versus control 1.27 ± 0.48 per field). Pepstatin also elevated LC3-II levels 2-fold but did not significantly enhance the effect of leupeptin when the two inhibitors were combined. As expected, virtually all AVs that accumulated after 24hr leupeptin treatment contained cathepsin D immunoreactivity, indicating that these structures were autolysosomes (Fig. 4E), confirming morphometric analysis showing 75% of the AVs appear to be autolysosomes.
To superimpose autophagy induction on inhibited lysosomal proteolysis, we grew additional neurons in EBSS, a glucose-containing medium that is devoid of amino acids and other growth factors. All EBSS-cultured neurons showed a rapid decrease in p70S6k expression relative to that in neurons grown in the presence of serum (Fig. 5A, B). LC3-II levels after starvation alone, were not detectably elevated and LC3-I levels were reduced after 24hr (11 ± 1%), suggesting that LC3-II turnover under these extreme induction conditions partially depletes LC3-I levels (Fig. 5C, D). By contrast, starvation in the presence of leupeptin raised LC3-II levels 20-fold after 24 hr (Fig. 5C, D). Starved neurons treated with pepstatin had no further effects on LC3 levels beyond those found in nutrient sufficient conditions or when combined with leupeptin. As seen after rapamycin-mediated autophagy induction, neurons starved of nutrients (EBSS) for 24 hr accumulated large vacuolar compartments that were devoid of most organellar material, corresponding to autolysosomes (Fig. 5E). By contrast, impeding degradation with leupeptin and pepstatin in the starved neurons caused granular material to accumulate in AVs (Fig. 5F) that were characterized ultrastructurally as primarily double-membrane-limited vesicles containing amorphous electron-dense or multilamellar material (Fig. 5G, H).
The microtubule-depolymerizing compound, vinblastine, disrupts cytoskeletal-dependent vesicular transport and subsequent fusion of autophagosomes with endosomal and lysosomal compartments (Marzella et al., 1980; Kovacs et al., 1982; Gordon and Seglen, 1988; Kochl et al., 2006). It has also recently been shown that in addition to blocking AV clearance by lysosomes, vinblastine induces autophagy (Kochl et al., 2006). Consistent with the latter observation, vinblastine (10μM) decreased p70 phosphorylation by 31 ± 12% at 6 hr and by 71 ± 7% at 24 hr (Fig. 6A,B). Vinblastine progressively raised LC3-II to levels well beyond those seen after inducing autophagy with rapamycin. LC3-II levels rose 128 ± 44% by 1 hr, 289 ± 93% after 6 hr and 444 ± 102% after 24 hr of rapamycin treatment (Fig. 6C, D), in the absence of a change in LC3-I levels. LC3 immunoreactivity also redistributed to abundant punctate structures in the cell body and neurites in the presence of vinblastine. Vinblastine treatment for 24 hr caused extensive AV accumulation and neuritic degeneration evident in immunolabeling studies with class III beta-tubulin antibody (Fig. 6H). By contrast, exposure of neurons to rapamycin or leupeptin had no discernable effects on microtubule integrity (Fig. 6F, G).
AVs accumulating within 1hr of vinblastine in DsRed-LC3 transfected neurons resembled endogenous LC3 immunolabeling after the same vinblastine treatment (Fig. 6I, left panel and inset). These AVs were autophagosomes which appeared red and lacked BODIPY-pepstatin (Fig 6I, right panel inset, arrow head) or autolysosomes that appear orange (red vesicle containing green BODIPY-pepstatin in the lumen; arrows) where AV-lysosome fusion has already occurred. By contrast, DsRed-LC3 vesicles containing BODIPY-pepstatin increased after 6hr of vinblastine (Fig. 6J, K) indicating that some extent of fusion with lysosomes occurs in the presence of vinblastine. To quantify the degree of delayed AV maturation caused by vinblastine, vesicular profiles of DsRed-LC3/BODIPY-pepstatin labeled neurons were quantified after rapamycin (10nM, 6hr) and vinblastine (10μM) for 1hr and 6hr (Figure 6L). DsRed-LC3 vesicles lacking any BODIPY-pepstatin were identified as autophagosomes (red; AP), while DsRed-LC3 positive vesicles containing BODIPY-pepstatin in the lumen were classified as autolysosomes (orange; AL). BODIPY-pepstatin-positive vesicles containing traces of DsRed-LC3 (yellow; LY) and BODIPY-pepstatin-only vesicles (green; LY) were characterized as lysosomes. Total vesicle counts per neuron were made for at least 15 cells under each condition and the fraction of each profile was reflected as a percentage of the total number of LC3 and BODIPY-pepstatin positive vesicles. Efficient AV maturation was found in rapapmycin-treated neurons, whose AV profiles were mostly autolysosomal (Fig. 6L, orange bar; 39.4%) and autophagosomes were rare (<1%). In contrast, vinblastine-mediated impairment of AV transport/fusion at 1hr increased the proportion of autophagosomes (red bar; 22.7%) while similar proportions of autolysosomes were maintained (39%). Extending vinblastine treatment for 6 hr did not increase the relative proportion of autophagosomes, but increased the percentage of autolysosomes (57.3%) and decreased the proportion of lysosomes. In contrast to rapamycin treatment, vinblastine reduced the proportion of lysosomes, which may indicate a delayed completion of proteolysis from autolysosome to lysosome. Initial maturation steps occur with vinblastine treatment, but the impaired fusion efficiency between AVs and lysosomes are such that incompletely degraded autolysosomes accumulate. At the ultrastructural level, the morphology of AVs that rapidly accumulated within the cell bodies (Fig. 7A) and neurites (Fig. 7A, inset) after 1 hr of vinblastine treatment corresponded frequently (44%) to autophagosomes (5.46 ± 0.86 per field, p<0,001 versus control 0.11 ± 0.15 per field), containing undigested uncompacted organellar material within single and double-membrane-limited vesicles (Figure 7A, middle panel). Lysosomotropic agents such as ammonium chloride (NH4Cl) neutralize acidic compartments and caused accumulation of morphologically similar autophagosomes by preventing autophagosome-lysosome fusion (Figure 7 A, bottom inset). By immunoelectron microscopy, cathepsin D was absent in these structures after short exposure to vinblastine, confirming that they were autophagosomes (Fig. 7 C). After 6 hr of vinblastine treatment, the AVs in neuronal cell bodies (Fig. 7B) and neurites (Fig. 7B, inset) were in the majority (63%) classified as autolysosomes (5.54 ± 1.04 per field, p<0.001 versus the numbers in control cells 1.27 ± 0.40 per field), which were single-membrane-limited vesicles containing amorphous electron-dense material and cathepsin D immunoreactivity (Fig. 7D). The change in morphology of AV intralumenal substrates between 1 and 6 hr suggested that, despite impaired fusion with lysosomes, limited hydrolysis of substrates is possible within autophagosomes, but is insufficient to degrade the inner limiting membrane of most autophagosomes.
In Alzheimer's disease, AVs with distinct morphologies accumulate in dystrophic neurites in massive numbers (Nixon et al., 2005; Yu et al., 2005). The most prevalent morphology consisted of single and double-membrane-limited vesicles with densely compacted amorphous or multilamellar contents (Fig. 8A). The majority of the compartments were labeled by immunogold-conjugated cathepsin D antibodies in immunoelectron microscopy analysis (Fig. 8B) (Nixon et al, 2005). Similar AV morphologies were seen in dystrophic neurites from PS/APP mice (Fig. 8C). The patterns within dystrophic neurites of AD brains were distinct from those that form in neurites and perikarya of treated primary cortical neurons after autophagy induction alone or after acute autophagosome-lysosome delivery impairment (Fig. 8D - H). By contrast, AV morphologies in AD strongly resembled the AV patterns generated in neurites and cell bodies of primary cortical neurons in which autophagosome proteolysis was inhibited or events required for efficient autophagosome-lysosome fusion were impaired but not entirely prevented (Fig.8 I - L). This morphologic pattern and AV accumulation were further accelerated when autophagy induction was superimposed.
Although the scarcity of AVs in the healthy brain (Mizushima et al., 2004; Nixon et al., 2005) has suggested a low activity of autophagy in neurons, we provide evidence that constitutive autophagy is relatively active in primary cortical neurons. Autophagosomes accumulated relatively rapidly when their clearance was prevented by blocking lysosomal proteolysis with cathepsin inhibitors. Considering the inhibition of lysosomal degradation did not induce autophagy, accumulating AVs may represent constitutive autophagosome formation in neurons under these conditions. The findings accord with our observation that intraventricular infusion of leupeptin in mice also induces a marked buildup of electron-dense autophagosomes and autolysosomes in neurons (Schmidt, Lee, Nixon, unpublished data). Our results also complement recent studies in which blocking autophagosome formation by deletion of the Atg5 or Atg7 genes caused accumulation of ubiquitinated proteins and eventual neurodegeneration, demonstrating that autophagy is both constitutive and essential for neuronal survival (Hara et al., 2006; Komatsu et al., 2006). In our study, the rapid accumulation of AVs within hours after protease inhibition in primary cortical neurons highlights an even greater basal level of autophagic activity than is suggested by the slow accumulation of ubiquitinated proteins over many days when autophagosome formation is prevented in vivo.
Of significant relevance to pathological states of autophagy, our data show that newly formed autophagosomes are normally eliminated efficiently in neurons by fusing with lysosomes, thereby avoiding a build-up of autophagic intermediates even when autophagy is strongly induced. The efficiency of this process in neurons accounts for earlier observations that AV intermediates in the autophagic pathway are relatively uncommon in normal brain tissue (Mizushima et al., 2004; Nixon et al., 2005). Neurons exhibited similar responses to strong autophagy induction by two different inhibitors of the mTOR kinase pathway: rapamycin, a direct inhibitor of the mTOR complex (Blommaart et al., 1995; Sabers et al., 1995), and nutrient deprivation, which inactivates the PI3K-AKT pathway leading to mTOR suppression (Barber et al., 2001; Vander Haar et al., 2007). Surprisingly, under maximal mTOR suppression, LC3-II-positive autophagosomes were only modestly increased in number, although AVs in later stages of digestion identified by electron microscopy were much more numerous. These vesicular profiles, which were strongly cathepsin-positive, correspond to autolysosomes (Lawrence and Brown, 1992; Eskelinen, 2005), which are larger than lysosomes, relatively electron-translucent and contain partially digested substrates. These observations indicate that, even under strong autophagy induction conditions, the fusion of autophagosomes with late endosomal and lysosomal compartments is highly efficient. Interestingly, in non-neuronal cells, such as fibroblasts and blastocysts (Tanida et al., 2004; Yu et al., 2005) acute autophagy induction does lead to greater transient accumulations of autophagosomes and larger LC3-II elevation than in neurons, underscoring cell type differences in the efficiency of AV clearance by lysosomes.
During AV clearance, fusion with a lysosome introduces into the autophagosome over 60 hydrolases with acidic pH optima and a membrane-spanning proton pump (V-ATPase) that lowers the intralumenal pH to levels required for optimal protease activation (Sun-Wada et al., 2003). Intact microtubules are needed for AV transport, which facilitate fusion between autophagosomes and late endosomal and lysosomal compartments (Kovacs et al., 1982; Kochl et al., 2006). As predicted, vinblastine, an inhibitor of microtubule assembly and of autophagosome transport to lysosomes, induced the rapid accumulation of autophagosomes and prevented the turnover of LC3-II associated with these compartments. The accumulation of immature AVs in neurites in these neurons within hours of vinblastine treatment (as also observed with protease inhibitors), accords with evidence for active ongoing autophagy in these processes. Earlier studies of growing neurons in culture have shown that autolysosome formation is active in neurites, as evidenced by the progressive acidification of retrogradely transported vesicles with AV characteristics (Overly et al., 1995). In our study, longer vinblastine treatment transformed large autophagosomes containing undigested organellar material into smaller (0.3 to 1.0 μm) vesicles. These vesicles retained some features of immature AVs (e.g. double membranes), but exhibited an amorphous dense intralumenal content, implying that limited proteolysis occurred within these compartments even though autophagosome maturation was incomplete. Although vinblastine indirectly reduces AV fusion with endosomal and lysosomal compartments by disrupting microtubule-dependent AV transport (Gordon and Seglen, 1988; Kochl et al., 2006), evidence that fusion itself is unaffected has been reported (Punnonen and Reunanen, 1990). Thus, residual stochastic fusion events may be sufficient to partially digest AVs, as some cathepsins are active over a broad pH range up to neutral pH (Turk et al., 1999; Pillay et al., 2002).
Impeding AV clearance either by inhibiting autophagosome-lysosome delivery or by blocking substrate proteolysis induced a relatively rapid accumulation of AVs with morphologies very similar to those seen in the AD brain and in the PS/APP mouse model of AD. These included large proportions of single and double-membrane limited vesicles with dense amorphous or granular content, contrasting sharply with the autolysosomes seen after strong autophagy induction in neurons with unimpaired AV clearance. These observations strongly support the hypothesis that profuse AV accumulation in AD neurons is reflective of impaired AV clearance from affected neurites in addition to any effects on autophagy induction. Defective axonal transport of vesicular cargoes back to the cell body has been proposed as one possible mechanism for organelle accumulation within axons in various metabolic and traumatic axonal injury states (Hollenbeck, 1993; Wang et al., 2006). Axonal swellings developing early in AD brain and in mouse models of AD (Stokin et al., 2005) have been attributed to axonal transport failure. In these settings, a variety of organelle types are usually present in the swellings. By contrast, AVs accumulate relatively selectively when proteolysis in autolysosomes is inhibited in vivo after leupeptin administration or in mice in which the genes for cathepsins D or cathepsins B and L are deleted (Bi et al., 1999; Shacka et al., 2007). These observations, while not inconsistent with a general defect in axonal transport, raise the additional possibility that a change in the properties of AVs may influence their axonal transport specifically, without necessarily altering general transport mechanisms. The transport kinetics of individual types of endocytic and autophagic vesicles in axons differ (Hollenbeck, 1993; Deinhardt et al., 2006) and an impairment in the maturation of AV subtypes could alter transport properties and favor AV accumulation. Interestingly, a number of recent studies on various lysosomal storage disorders have also found evidence of impaired AV clearance (Kiselyov et al., 2007; Settembre et al., 2008).
Our evidence showing that autophagy induction in the face of impaired AV clearance exacerbates AV pathology has possible implications for autophagy modulation as a therapeutic strategy in neurodegenerative diseases. Recent evidence suggests that autophagy enhancement is a promising therapeutic strategy against the toxicity of misfolded proteins in age-related neurodegenerative diseases (Rubinsztein et al., 2007) . In diseases where the autophagic pathway may be normal or sluggish but not defective, as seems to be the case in Huntington disease models (Rubinsztein et al., 2005), pharmacologically enhancing the sequestration of misfolded proteins would be expected to promote more rapid digestion and clearance of toxic protein with therapeutic benefits, as has been observed (Ravikumar et al., 2004; Sarkar et al., 2007). In AD where autophagosome clearance may be impaired, strongly inducing autophagosome formation in AD may exacerbate an already massive neuronal buildup of “intermediate” autophagic compartments, some of which are able to generate Aβ (Yu et al., 2005) and possibly other toxic metabolites. Both positive and negative effects of autophagy have been proposed in neuronal function (Chu, 2006), and the future delineation of specific (primary) and non-specific (secondary) autophagic dysfunction in different neurodegenerative diseases will provide a better insight into the differences in their underlying pathology. From results obtained in this study, we propose that therapeutic modulation of autophagy in AD may need to improve the efficiency of autophagosome maturation or eventual substrate digestion by lysosomes.
The authors gratefully acknowledge the assistance of Corrinne Peterhoff, Nicole Piorkowski and Heather Braunstein, at NKI in preparing this manuscript, along with Dr. Ana Maria Cuervo at Albert Einstein College of Medicine in New York, for her feedback on the manuscript. We also thank Dr Panaiyur S. Mohan at NKI for supplying the LC3 antibody used and Dr. John S. Young from the Dept. of Pharmacology at Oxford for his assistance with confocal microscopy imaging. Research by the authors was supported by grants from the National Institute on Aging (AG 017617), the Alzheimer's Association, and the Wellcome Trust (B. Boland).