Accumulation of amyloid-like Aβ_1–42 in AEL (autophagy–endosomal–lysosomal) vesicles: potential implications for plaque biogenesis

Fly strains and Aβ expression

Drosophila melanogaster were raised using standard methods. Human Aβ_1–40 or Aβ_1–42 was expressed in Drosophila central nervous system using the binary Gal4-UAS technique as illustrated in Supplementary Figures S1A and S1B (at http://www.asnneuro.org/an/006/an006e139add.htm). Fly strains used in this work (including 3.1 kb Gad1-Gal4, 7.4 kb Cha-Gal4, Elav-GeneSwitch-Gal4, UAS-Aβ_1–40, UAS-Aβ_1–42, UAS-GFP^S65T, UAS-GFP-Atg8a, UAS-Atg5^RNAi, UAS-Atg12^RNAi, Atg1^∆3D) were previously described (Ling et al., 2009; Ling and Salvaterra, 2011b). Additional fly strains are UAS-mCherry-Atg8a (Chang and Neufeld, 2009) and UAS-mRFP-Rab4 (obtained from the Bloomington Drosophila Stock Center).

Neuron culture and measurement of Aβ secretion

Embryos at late gastrula stage were collected from Elav-GS-Gal4:UAS-GFP flies that incorporate either the UAS-Aβ_1–40 or UAS-Aβ_1–42 transgene. Embryonic cells were dissociated and cultured as described (Salvaterra et al., 1998). Neurons were differentiated for 5 days at 25°C in Schneider's Drosophila medium (Invitrogen) supplemented with 10% FBS, 0.3 μg/ml insulin, 100 U/ml penicillin and 100 μg/ml streptomycin. Culture medium was then replaced by medium with reduced serum (1%) containing 0.1 μM RU486 (Sigma) to induce Aβ expression for 5 days. Parallel cultures with no RU486 inducer served as a negative control. Culture medium was collected for measurement of extracellular Aβ secretion. Cells were harvested and lysed in ice-cold RIPA/SDS buffer [50 mM Tris/HCl (pH 8.0), 0.5% sodium deoxycholate, 1% Triton X-100, 150 mM NaCl and 1% SDS] plus protease inhibitors (Roche). Aβ in culture medium or cell lysate was immunoprecipitated using the anti-Aβ antibody 6E10 (Covance) and Protein A/G plus agarose (Santa Cruz Biotechnology). The immunoprecipitated proteins were eluted and separated by electrophoresis through a 4–20% Tris-glycine gel (Invitrogen), transferred to PVDF membranes and detected using anti-Aβ 6E10 as the primary antibody and HRP (horseradish peroxidase)-conjugated chicken anti-mouse IgG (Santa Cruz Biotechnology) secondary antibody following incubation with SuperSignal West Pico chemiluminescent substrate (Pierce). Densitometric analysis of Western blot bands was performed using ImageJ (NIH). The densitometry of Aβ bands were normalized by the immunoprecipitation antibody bands at 50 kDa (Figures 2C and 2E). For correction of the Aβ species-specific sensitivity to the primary antibody, 1 pmol of synthetic Aβ_1–40 or Aβ_1–42 (California Peptide Research) was added to the negative control samples as an Aβ species-specific calibrator.

Cathepsin D activity assay

Frozen heads from 30 adult females were homogenized in 100 μl of ice-cold buffer [10 mM Tris (pH 7.5), 5 mM EDTA and 0.25 M sucrose] and centrifuged at 20000 g for 10 min at 4°C to remove debris, nuclei and large mitochondria. The supernatant was centrifuged at 200000 g for 45 min to collect microsomal membranes [i.e. lysosomes, autolysosomes, plasma membranes, ER (endoplasmic reticulum), Golgi, etc.]. The pellet was dissolved in 100 μl of 0.1M sodium acetate buffer (pH 5.0) containing 0.1% Triton X-100. CathD (cathepsin D) activity was measured as described (Yasuda et al., 1999) using a fluorescently labeled CathD substrate. Samples were incubated at 35°C in sodium acetate (pH 3.5) for up to 3 h and fluorescence was measured using a SpectraMax microplate reader at 328 nm (excitation) and 393 nm (emission). Enzyme activity was averaged from multiple independently prepared replicates and presented as arbitrary relative fluorescence units/h per μg of protein. Enzyme assays were linear with respect to sample and incubation time. Total proteins were measured using a BCA assay (Pierce) for data normalization.

Western blotting assay

Fly heads from 25-day-old females were homogenized in RIPA/SDS buffer [50 mM Tris/HCl (pH 8.0), 0.5% sodium deoxycholate, 1% Triton X-100, 150 mM NaCl and 1% SDS] with protease inhibitors (Roche Complete). Samples were incubated for 1 h at 4°C and centrifuged at 12000 g for 20 min. The supernatant contained RIPA/SDS-soluble Aβ_1–42. Pellets were washed with RIPA buffer, homogenized in 70% FA (formic acid), incubated 1 h at 4°C and centrifuged at 12000 g for 20 min. The supernatant was collected, dried using a SpeedVac centrifuge (Savant) and dissolved in DMSO to obtain FA-soluble Aβ_1–42. Proteins were separated on 4–20% polyacrylamide gradient gels (Invitrogen), transferred on to PVDF membranes (Bio-Rad Laboratories), detected by anti-Aβ 6E10 antibody staining (Covance) and HRP-conjugated secondary antibody, and visualized with SuperSignal West Pico chemiluminescent substrate (Pierce) to expose X-ray film. Densitometry of Aβ or loading control bands (40 kDa unspecific protein) was performed using Bio-Rad Quantity-One software.

Immunohistochemistry

Dissected brains were fixed in 4% paraformaldehyde at 4°C overnight. For immunostaining using anti-Aβ antibody 4G8 (Covance) or rabbit polyclonal anti-Aβ_1–42 antibody ab12267 (AbCam), brains were pre-treated with 70% FA for 10 min. For immunostaining using aggregated-Aβ-specific antibody 7A1a (New England Rare Reagents), FA pre-treatment was not included. Primary antibody immunostaining was detected using Alexa Fluor-555-conjugated secondary antibody (Invitrogen). For immunostaining with anti-Rab5 antibody (AbCam), the secondary antibody used for detection was Alexa Fluor-647-conjugated goat anti-rabbit IgG (Invitrogen). Samples were observed and images collected using confocal microscopy (Zeiss LSM 510).

Congo Red histochemical staining

Fly heads were fixed in 4% paraformaldehyde at 4°C for 24 h then washed thoroughly in PBS followed by infiltration with graded concentrations of sucrose. Cryosections were stained with Congo Red (Sigma) following a published procedure (Wilcock et al., 2006). Microscopy images were collected using a 60× objective for both bright-field and dark-field illumination under polarized light.

Fluorescent Congo Red, LysoTracker and CellMask imaging

For Congo Red fluorescence imaging, the brains were fixed and washed as above then stained with 0.2 mg/ml Congo Red in PBS for 10 min. The Congo Red solution was freshly prepared and filtered through a 0.2 μm filter before use. After PBS washing, brains were observed using confocal microscopy. Congo Red fluorescence was detected as described (Wiesehan et al., 2003). LysoTracker red (Molecular Probes) staining was performed as previously described (Ling et al., 2009). For the CellMask plasma membrane staining, whole brains were first immunostained using anti-Aβ 4G8 (Covance) and Alexa Fluor-555-conjugated goat anti-mouse IgG (Invitrogen) secondary antibody, then washed in PBS, followed by a 10 min incubation in PBS containing 5 μg/ml CellMask™ Deep Red plasma membrane stain (Invitrogen). The brains were not permeabilized with detergent prior to the CellMask staining. After 5×PBS washing, brains were mounted and observed by confocal microscopy.

Fluorescence and electron microscopy

Expression of UAS-fluorescent protein transgenes (cytosolic GFP, mCherry-Atg8a, GFP-Atg8a, and mRFP-Rab4) was controlled by a Gal4 driver and fluorescence detected using confocal microscopy on adult brains fixed in PBS with 4% formaldehyde for 30 min, followed by thorough washing in PBS. Electron microscopy was performed as previously described (Ling et al., 2009) using an FEI Tecnai transmission electron microscope. Independent observations from three to five animals were performed for each experimental condition.

Co-localization analysis

Transgenic GFP-Atg8a and mRFP-Rab4 were expressed in neurons of control, Aβ_1–40 and Aβ_1–42 flies. The brains from age-matched female adults were dissected, fixed and observed using confocal microscopy. Kenyon cells from Drosophila mushroom bodies were selected as the area of interest due to high neuronal density in this brain region. Z-stack images with multiple channels (green for GFP-Atg8a and red for mRFP-Rab4) were collected using a 63× (NA=1.2) water-immersion objective. At least ten optical sections with z-spacing of 0.5 μm were acquired. Image collection from all samples used identical microscopy conditions including the size of the pinhole, optimized contrast and detector gain. Two independent z-stacks were collected from each brain sample and at least three independent brains were observed for each experimental condition. Co-localization assays were performed using ImageJ (NIH) combined with the Intensity Correlation Analysis plugin (Li et al., 2004). Mander's overlap coefficient was used to quantify co-localization of GFP-Atg8a and mRFP-Rab4 puncta. Mander's overlap coefficient ranges from 0 (no co-localization) to 1 (100% co-localization).

RT–qPCR (reverse transcription–quantitative PCR)

RT-qPCR was performed as described (Ling and Salvaterra, 2011c). The primers for Aβ_1–40 and Aβ_1–42 transgenes were 5′-ATGAGTCCAATGATTGCACCT-3′ and 5′-AGACTTTGCATCTGGCTGCTA-3′. Ten reference genes [14-3-3ε, Appl, Cyp1, Elav, eIF-1A, l (3)02640, Rap2l, RpL32, Su (Tpl) and αTub84B] were measured for data-specific normalization (Ling and Salvaterra, 2011c). C_T values from real-time PCR were analyzed using a custom SASqPCR program (Ling, 2012). Mean normalized expression ratios were calculated as described (Ling et al., 2012).

Statistical analysis

Image analyses, including densitometric quantification of immunoprecipitation/Western blotting (Figures 2D and 2F), mean signal intensity of anti-Aβ immunostaining (Figure 1C), quantification of the mRFP-Rab4 objects (Figure 5D) and co-localization assay (Figure 5E), were performed using ImageJ (1.45s, NIH). Densitometric analysis of non-immunoprecipitation Western blotting was performed using Bio-Rad Quantity-One (Figure 5). The data analyzer was blinded to image identities relevant to experimental conditions. The sample sizes for biological replicates (n) are provided in relevant Figure legends. Data are presented as means±S.D. or S.E.M. as indicated. RT–qPCR data analysis for measurement of Aβ transcript levels (Figures 1D and 6D) was as described above. The normalized expression ratios are presented as means±S.E.M. For all statistical comparisons, two-tailed P values were obtained by Student's t test or ANOVA [corrected for multiple comparisons (Bonferoni) using GraphPad Prism5]. The α-level for all tests was set at 0.05.

Aβ_1–42 appears to be selectively associated with membrane

Human Aβ_1–40 or Aβ_1–42, fused to a rat pre-proenkephalin secretory signal sequence (Finelli et al., 2004), were expressed separately in Drosophila neurons using the Gal4:UAS bipartite gene expression technique. Targeted neurons were additionally labeled by co-expression of soluble GFP which initially fills the cytosol and neurites and is detected by fluorescence microscopy (see Supplementary Figure S1). Anti-Aβ antibody 4G8 immunostaining was applied to whole brains to compare the initial accumulation and distribution of Aβ_1–40 and Aβ_1–42 peptide in brain region-matched samples from relatively young specimens (1-day-old, relative to eclosion, Figures 1A and 1B). The 4G8 antibody targets the 17th–24th amino acids common to both Aβ_1–40 and Aβ_1–42. The relative staining intensity appears significantly higher for Aβ_1–42 than Aβ_1–40 (Figure 1C). However, transcript levels of Aβ_1–42 mRNA detected by RT–qPCR are lower than those of Aβ_1–40 (Figure 1D). The distribution of Aβ_1–42 staining is strikingly different compared with Aβ_1–40 staining. Aβ_1–42 staining appears to outline cell somas (Figure 1B). In brains taken from older specimens (16-day-old) Aβ_1–40 staining appears to be mostly focal (Figure 1E) and remains less intense than Aβ_1–42 (Figures 1F and 1G). In these older samples, Aβ_1–42 staining still appears to be associated with some somal membranes, but staining is also evident in intracellular foci as well as in some likely neurites (Figures 1F and 1G). These intracellular foci were previously identified as dysfunctional autophagy–lysosomal vesicles and their number and size increases with age specifically in Aβ_1–42-expressing samples (Ling et al., 2009). Our results suggest that Aβ_1–42 may preferentially associate with various membranous structures, including plasma membranes, whereas Aβ_1–40 does not.

Neurons preferentially secrete Aβ_1–40 but retain Aβ_1–42

The C-terminus of Aβ_1–42 has an additional isoleucine and alanine relative to Aβ_1–40 that increases its hydrophobicity and membrane association (Marchesi, 2005). This increased membrane association may contribute to its greater accumulation within neurons by interfering with efficient secretion following biosynthesis through the secretory pathway. To obtain evidence for this possibility, we double-stained fly brains expressing Aβ with anti-Aβ 4G8 antibody and CellMask, a plasma membrane-specific fluorescent dye. In 1-day-old Aβ_1–42 samples we observe extensive co-localization of antibody staining with CellMask staining (Figure 2B) suggesting an association of Aβ_1–42 with plasma membranes. In contrast, Aβ_1–40 samples exhibit only limited co-localization of a few plasma membrane foci along with some intracellular foci (Figure 2A). A close examination of Aβ_1–40 staining reveals that additional Aβ_1–40-positive signal is localized in regions not coincident with CellMask, but rather present in adjacent areas that may be extracellular spaces (Figure 2A, inserts 1 and 2). Aβ_1–42 staining in contrast is not found in equivalent areas (Figure 2B, inserts 3 and 4) suggesting that Aβ_1–40, but not Aβ_1–42, may be more amenable to secretion into extracellular spaces. To test this possibility directly, we expressed Aβ_1–42 or Aβ_1–40 in primary cultured Drosophila neurons using a drug-inducible system to conditionally control expression (Supplementary Figures S1C and S1D). Temporal control of expression should minimize confounding factors that might result from specific neuronal toxicity of Aβ_1–42. Quantification of Aβ in the culture medium or within neurons indicates that Aβ_1–40 is preferentially recovered from culture medium while Aβ_1–42 accumulates more prominently within neurons (Figures 2C–2F). These results support the proposal that neurons preferentially retain Aβ_1–42, likely a consequence of its higher membrane affinity, but preferentially release Aβ_1–40.

Movie 1.

Vesicles accumulating punctate GFP and Aβ_1–42 have endosomal properties

A prominent morphological change induced by Aβ_1–42, but not Aβ_1–40, in our Drosophila model is a redistribution of the normal homogeneous cytosolic GFP fluorescence into punctate structures (Figures 3A–3C and Supplementary Movie S1 at http://www.asnneuro.org/an/006/an006e139add.htm). Our previous work established that these puncta are derived from an age-dependent autophagic sequestration of cytosolic GFP which is resistant to degradation by lysosomal hydrolases and is thus an indicator of long-lived dysfunctional autophagy–lysosomal vesicles (Ling et al., 2009). In many targeted neurons the contents of these vesicles leaks out into cytoplasm where it appears to initiate intraneuronal necrosis and a decrease in the fluorescence intensity of non-vesicular cytosolic GFP, possibly by reactivation of lysosomal hydrolases. This process increases the contrast of non-degraded GFP fluorescence contained within the vesicles relative to surrounding cytosolic GFP fluorescence (Ling et al., 2009; Ling and Salvaterra, 2011b).

Since Aβ_1–42 is initially associated with plasma membrane and the autophagy and endosomal pathways converge (Eskelinen, 2005), we examined the possible contribution of endosomal involvement in the vesicle compartment recognized by punctuate GFP fluorescence and Aβ_1–42 staining (Figure 1G). We first confirmed that GFP puncta substantially co-localize with LysoTracker red staining (a lysosomal marker indicating acidic pH, Figure 3D) and transgenic mCherry-labeled autophagy-specific gene 8a protein (mCherry-Atg8a, Figure 3E) identifying the contributions of autophagy and lysosomal vesicle fusion. In addition we observe that most of the GFP puncta also co-localize with transgenic mRFP-labeled Rab4 expression (Figure 3F) as well as anti-Rab5-specific antibody immunostaining (Figure 3G). Rab4 and Rab5 are endosomal markers (Sönnichsen et al., 2000) and these data thus identify the Aβ_1–42-induced GFP puncta as dysfunctional AEL vesicles. Expression of cytosolic GFP alone or in combination with Aβ_1–40, in contrast with the Aβ_1–42 expression, does not result in accumulation of significant numbers of GFP puncta (Figures 3A and 3B and Supplementary Movie S1). Our results thus suggest that Aβ_1–42, initially associated with plasma membrane, may be transported in part to AEL vesicles through the convergence of endosomal, autophagy and lysosomal structures. The absence of GFP catabolism, evident from its continued fluorescence in the AEL vesicles, along with decreased or absent turnover of the vesicles, appears to be a specific consequence of neuronal Aβ_1–42 expression, but the specific catabolic inhibitory mechanism remains unknown.

Inefficient vesicle turnover promotes continued fusion of AEL vesicles

Abnormal accumulation of autophagic and endosomal vesicles relevant to AD neuropathology has been postulated to result from a blockage of vesicle fusion with lysosomes, resulting in a failure to acquire catabolic enzymes necessary for cargo degradation (Funk and Kuret, 2012). We co-expressed GFP-Atg8a and mRFP-Rab4 in control, Aβ_1–40 or Aβ_1–42 animals to test if the decreased vesicle turnover in Aβ_1–42-expressing neurons affects the convergence of autophagy and endosomal markers. Compared with age- and brain region-matched control samples with no Aβ expression (Figure 4A) or Aβ_1–40 samples (Figure 4B), we observe enhanced accumulation of autophagic and endosomal vesicles in Aβ_1–42 samples (Figures 4C and 4D). Since vesicle turnover is efficient in healthy neurons, this increased accumulation of AEL vesicles in Aβ_1–42 samples suggests their turnover is slower and that they maintain an extended duration in cytosol. Co-localization of autophagic and endosomal markers is significantly greater in brains from Aβ_1–42 flies compared with either control or Aβ_1–40 flies (Figure 4E), suggesting that the decreased vesicle turnover in Aβ_1–42 samples may actually promote fusion between autophagic and endocytic vesicles. In addition, we frequently observe complex intermediate structures that may be a manifestation of different stages of fusion between the diverse types of AEL vesicles in brain tissue in Aβ_1–42 samples (Figures 4F and 4I). Some of these appear to be extremely large AEL vesicles (Figures 4J and 4M) that may result from a combination of extended vesicle duration and multiple vesicle fusion among dysfunctional AEL vesicles. In contrast, large AEL vesicles are absent from age-matched control or Aβ_1–40 samples (Supplementary Figure S2 at http://www.asnneuro.org/an/006/an006e139add.htm).

Decreasing AEL function results in decreased Aβ_1–42 accumulation

We decreased functional autophagy activity by expressing small interfering RNA transgenes (UAS-RNAi) targeting either autophagy-specific gene 5 (Atg5^RNAi) or 12 (Atg12^RNAi) in Aβ_1–42-expressing neurons. Both RNAi genotypes exhibit a similar reduction in CathD activity (the major lysosomal aspartyl protease; Figure 5A) or the number of LysoTracker-positive structures (Supplementary Figure S3 at http://www.asnneuro.org/an/006/an006e139add.htm) suggesting successful reduction in both the function and catabolic capacity of convergent AEL vesicles. Expression of either RNAi transgene also results in decreased RIPA/SDS-soluble or FA-soluble Aβ_1–42 accumulation (Figures 5B and 5C). Since the RNAi genotypes do not have significantly different levels of Aβ_1–42 transcripts relative to control samples, the reduction in Aβ_1–42 accumulation is likely to be post-transcriptional. The RNAi-mediated decrease in AEL function is limited to Aβ_1–42-expressing neurons. We additionally used a conventional loss-of-function allele for the autophagy-specific kinase 1 gene (Atg1^∆3D) that decreases AEL activity in all cell types (Scott et al., 2004). Heterozygous Atg1^∆3D animals exhibit decreased CathD activity in fly heads as expected (Figure 5E) and also exhibit a reduction in accumulation of Aβ_1–42 (Figure 5F). These data suggest the interesting possibility that AEL vesicle formation may be necessary for effective accumulation of Aβ_1–42 in Drosophila neurons. It also seems likely that Aβ_1–42 may in turn participate in development of the AEL dysfunctional vesicles.

Aβ_1–42 within AEL vesicles is recognized with aggregate-specific anti-Aβ antibodies

Aβ_1–42 is an aggregate-prone peptide and its formation into higher-ordered aggregation states is thought to be a key step in development of amyloid plaques in AD. To examine the relationship between dysfunctional AEL vesicles and Aβ_1–42 amyloid-like aggregation, we performed immuno-staining of fly brains using anti-Aβ antibodies that recognize forms of Aβ. Immunostaining with anti-Aβ 4G8 antibody, which does not distinguish between diffuse and aggregated Aβ_1–42, stains GFP puncta as well as plasma membranes and other types of fibrous-appearing intracellular Aβ_1–42 structures (Figures 1G and 6A). However, immunostaining with aggregate-specific antibody 7A1a (Zhou et al., 2012) primarily co-localizes with GFP puncta (Figure 6B). A comparison of 4G8 and 7A1a immunostaining thus suggests that highly aggregated forms of Aβ_1–42 may be limited to the dysfunctional AEL vesicles. This result was further confirmed by immunostaining with a different anti-Aβ_1–42 antibody (AbCam, ab12267) with reportedly high affinity for plaque-like aggregations of Aβ_1-42 (http://www.abcam.com/beta-amyloid-1-42-antibody-ab12267.html). The ab12267 immunostaining co-localizes well with many of the AEL vesicles identified by expression of transgenic GFP-Atg8a protein (Figure 6C). Our results thus suggest that aggregated forms of Aβ_1–42 with potential amyloid-like properties are selectively associated with the dysfunctional AEL vesicles. The non-AEL-localized Aβ_1–42 is in a non-aggregated state, a result that suggests the vesicles themselves may contribute to amyloid-like aggregation.

AEL vesicles have typical amyloid plaque-like features

This focal aggregation of Aβ_1–42 specifically within dysfunctional AEL vesicles could be a source of amyloid plaques. To test this, we performed Congo Red histochemical staining on brain sections. Congo Red staining that also exhibits apple-green birefringence is considered to be a ‘gold standard’ for identification of amyloid plaques in human tissue (Sipe et al., 2010). Brain tissue from Aβ_1–42 flies, but not age-matched control (Figure 6D) or Aβ_1–40 flies (Figure 6E), exhibits congophilic staining with typical apple-green birefringence when observed under polarized light microscopy (Figures 6F and 6G) suggesting an Aβ_1–42-specific amyloid-type deposition. In addition, the congophilic staining appears to be discrete (Figure 6G) and thus consistent with localization restricted within AEL vesicles. To confirm the association between amyloid formation and AEL vesicles, we applied fluorescent Congo Red staining followed by confocal microscopy as described (Wiesehan et al., 2003). Congo Red fluorescence co-localizes in part with the punctate redistribution of transgenic GFP-Atg8a in Aβ_1–42 but not Aβ_1–40-targeted neurons (Figures 6H and 6I). These observations support the possibility that AEL vesicles containing aggregated Aβ_1–42 deposits could potentially be a source of extracellular amyloid plaques.

AEL vesicles may relocate from intra- to extra-cellular spaces

How could amyloid-containing intracellular AEL vesicles relocate to extracellular spaces? Our previous work showed that compromised AEL vesicles participate in a necrotic-type neurodegeneration in Drosophila (Ling et al., 2009; Ling and Salvaterra, 2011b). Intriguingly, nearly all cytosolic GFP fluorescence in some dying neurons eventually disappears; however, some fluorescent dysfunctional AEL vesicles remain (Figure 7A). This implies that the vesicles containing aggregated amyloid-like forms of Aβ_1–42 could remain in brain tissue even after the complete necrosis of their host neuron. Consistent with this possibility, we observe clusters of AEL vesicles in extracellular spaces (Figures 7B–7E) that conceivably could have been localized in a now completely degenerated neuron.

The delayed or absent turnover of dysfunctional AEL vesicles, in conjunction with cumulative fusion of additional autophagy and/or endosomal vesicles, also apparently results in the formation of extremely large vesicles which we routinely observe in Aβ_1–42-expressing samples (Figures 7J–7M), but not in age-matched control or Aβ_1–40 samples (data not shown). In some cases, these extremely large vesicles appear to be coincident with a process of tortuous infolding of the plasma membrane of their host neuron (Figures 7F and 7G). This infolding process could also conceivably lead to an eventual complete separation of the enlarged vesicles from their resident neuron (Figures 7H and 7I), even in the absence of any neurodegenerative changes in nearby neurons (Figures 7J–7M) and thus result in their appearance in extracellular spaces Similar morphological data is also apparent in very old Drosophila even in the absence of Aβ_1–42 expression and may thus be a consequence of a normal aging process (Ling and Salvaterra, 2011b).

Levels of Aβ_1–42 versus Aβ_1–40

The higher Aβ_1–42 protein immunocytochemical staining we observe in young fly brains compared with Aβ_1–40 (Figure 1C) is surprising in light of expression of similar transcript levels for these transgenes (Figure 1D). Although this could be a result of technical limitations of whole-brain staining, another possible explanation is that preferential membrane interaction of Aβ_1–42 may render it resistant to degradation. This possibility has been observed in other systems (Knauer et al., 1992; Burdick et al., 1997; Ling et al., 2009). Our results are also broadly in accordance with AD as well as some mammalian AD models where neurons preferentially produce Aβ_1–40 from APP proteolysis (Hartmann et al., 1997), but paradoxically accumulate higher levels of intraneuronal Aβ_1–42 (Gouras et al., 2000, 2005; Laferla et al., 2007). Similar differential accumulation levels have also been previously reported in Drosophila photoreceptor neurons (Finelli et al., 2004). The Drosophila model thus potentially recapitulates a key feature of AD, the predominant intracellular accumulation of Aβ_1–42 over Aβ_1–40. Interestingly, this phenotype is not dependent on amyloidogenic proteolytic processing of APP.

Recycling endocytosis may contribute to accumulation of Aβ_1–42 in AEL vesicles

Our previous work emphasized a prominent role for dysfunctional autophagy–lysosomal vesicles in Aβ_1–42 accumulation as well as age-dependent pathogenesis (Ling et al., 2009; Ling and Salvaterra, 2011b). Here we extend these observations by showing that the dysfunctional vesicles also express endosomal markers and thus represent convergent structures of autophagy, endosomal and lysosomal vesicle fusions.

A key question is how endosomal vesicles participate in the specific accumulation of Aβ_1–42 in the AEL vesicles. We can rule out AEL-dependent processing of APP since the only proteolytic processing necessary for Aβ_1–42 production in Drosophila is removal of the secretory signal peptide during trafficking through the secretory pathway. The complete removal of signal peptide from both Aβ_1–40 and Aβ_1–42 in Drosophila (Iijima et al., 2004) confirms its successful transit and processing through the secretory pathway. Only Aβ_1–42, however, associates with plasma membrane, especially at early ages, and specifically accumulates within dysfunctional AEL vesicles suggesting that recycling endocytosis of plasma membrane associated Aβ_1–42 followed by normal endosomal trafficking may at least in part contribute to its accumulation within AEL vesicles. Additional Aβ_1–42 may also accumulate within AEL vesicles directly as a result of autophagy. This may occur by virtue Aβ_1–42 association with non-endosomal membrane structures, including mitochondria or non-AEL-related vesicles that are targeted to autophagosomes as part of normal neuronal homoeostatic mechanisms. Non-membrane-associated intracellular Aβ_1–42 aggregates could also conceivably be targeted directly to the autophagy pathway, but our antibody staining only detects highly aggregated forms of Aβ_1–42 when they are already incorporated within AEL vesicles, especially in older samples. The reduction in Aβ_1–42 accumulation we observe following genetic down-regulation of autophagy activity is thus likely to be a consequence of reduced convergence of autophagy and endosomal vesicles. Our genetic results further suggest that AEL vesicles may be necessary for optimal Aβ_1–42 accumulation in Drosophila, a somewhat counterintuitive proposal given the catabolic nature of normal AEL function. We note, however, that the decrease in Aβ_1–42 accumulation is entirely consistent with our previous genetic observations that autophagy reduction in Drosophila lengthens lifespan and reduces the rate of age-dependent neurological deficits while pharmacologically increasing autophagy has an opposite effect in Aβ_1–42-expressing animals (Ling et al., 2009). Dysfunctional AEL processes may thus not only be necessary for optimal and preferential Aβ_1–42 accumulation, but also may be necessary for subsequent neurodegenerative mechanism(s) in Drosophila.

The relative importance of autophagy, endosomal or lysosomal vesicle trafficking for Aβ_1–42 accumulation is unknown; however, future genetic studies using endosomal and autophagy loss-of-function alleles in various combinations could potentially establish this. A recent study has also demonstrated that autophagy inhibition results in decreased Aβ accumulation in mammalian neurons (Nilsson et al., 2013), whereas other studies have found potentially conflicting results (Pickford et al., 2008; Boland et al., 2010). Many of these differences could be a result of various experimental details such as species differences in the type of APP being expressed, the particular types of AEL stimulation or inhibition used, confounding factors related to unspecified amyloidogenic APP processing (i.e. whether Aβ_1–42 or Aβ_1–40 predominates) or even the age and stage of neuronal compromise in the cells or neurons being studied. In AD, as well as the majority of mammalian AD models, Aβ must be generated by amyloidogenic proteolytic processing of APP, a process that reportedly occurs at both plasma membrane (Armstrong, 1998; Takahashi et al., 2002b; Marchesi, 2005) as well as intracellular AEL vesicles (Cataldo et al., 2004; Yu et al., 2005). It remains uncertain if either site predominates quantitatively (Haass et al., 2012), but both intracellular and extracellular Aβ are believed to be important for Aβ proteotoxicity as well as intracellular accumulation (Gouras et al., 2005; Mohamed and Posse de Chaves, 2011). Importantly, intracellular Aβ accumulation appears to precede extracellular Aβ deposition as well as plaque formation in some models (Gyure et al., 2001; Knobloch et al., 2007) and Aβ_1–42 is the primary form of peptide accumulating within mammalian neurons (Laferla et al., 2007), results consistent with our observations in Drosophila.

Dysfunctional AEL vesicles may be the source of plaque-like structures

Our previously data (Ling et al., 2009; Ling and Salvaterra, 2011d), as well as data presented here suggest a potential AEL-dependent mechanism for extracellular plaque-like formation in Aβ_1–42-expressing Drosophila. First, APP amyloidogenic proteolysis is not necessary. Both recycling endocytosis as well as autophagy could provide the necessary routes for Aβ_1–42 accumulation within AEL vesicle compartments and the relative selectivity of Aβ_1–42 over Aβ_1–40 could result from its greater membrane association. In young animals, the Aβ_1–42-containing AEL vesicles are capable of cargo digestion and would thus turnover at an appreciable rate. Aged animals, however, have a decreased efficiency for cargo degradation as a consequence of normal aging processes (Cuervo et al., 2005). In Aβ_1–42-expressing Drosophila neurons the dysfunctional AEL vesicles persist for long durations, continue to fuse with additional endocytic, autophagic and lysosomal vesicles and can become extremely large. Since Aβ_1–42 is also present in non-AEL locations in a non-amyloid form in Drosophila neurons, in vivo accumulation of Aβ_1–42 alone does not appear to be sufficient for amyloid formation. Amyloid formation may rather be dependent on the acidic and lipidic microenvironment of AEL vesicles, conditions shown to be favorable for formation of toxic aggregated forms of Aβ in mammalian neurons (Su and Chang, 2001). We have previously shown that some AEL vesicles leak their contents into cytoplasm (Ling et al., 2009). This process might conceivably create a non-AEL acidic and lipidic microenvironment in cytoplasm that could also convert non-amyloid Aβ_1–42 into an amyloid-like form that is subsequently reincorporated into AEL vesicles. Finally, dysfunctional AEL vesicles containing amyloid-like Aβ_1–42 would be relocated to extracellular spaces by two distinct mechanisms: neurodegeneration itself or separation of large dysfunctional AEL vesicles through a process of plasma membrane infolding.

The origin of extracellular senile plaques in AD is still a matter of debate. A prevalent view holds that they are formed by an autonomous condensation of extracellular Aβ peptides released from plasma membrane proteolysis of APP (Armstrong, 1998; Fiala, 2007). Alternative views emphasize an intracellular origin, including generation of Aβ directly by AEL vesicles (Glabe, 2001). Our proposed AEL-relocalization model in Drosophila may also have relevance for AD and mammalian AD models (Takahashi et al., 2002a; D’Andrea et al., 2003). The congophilic staining in Drosophila is discrete and consistent with amyloid restriction to a vesicular structure. Similar microdeposits have been observed in aged transgenic mouse brains expressing mutant APP (Takahashi et al., 2004; Stokin et al., 2005). The long duration and continuing fusion of membrane vesicles involved in AEL formation could easily result in production of not only large-sized amyloid plaques, but also a diversity of plaques sizes with spherical shapes, a common feature of AD and other AD models including our Drosophila AD model (Fiala, 2007; Ling et al., 2009; Ling and Salvaterra, 2011b). This proposed plaque model is also consistent with lysosomal materials, damaged organelles and other intracellular contents associated with AD plaques (Suzuki and Terry, 1967; Cataldo and Nixon, 1990; Fiala, 2007). It could obviate the difficulties in reconciling the different proportions of Aβ species found in diffuse, primitive and mature dense-core plaques (Fiala, 2007) believed to represent different stages of extracellular plaque maturation (Armstrong, 1998) with observations that senile plaques themselves contain primarily Aβ_1–42. The model could even explain why Aβ is constitutively produced in human brain and peripheral tissues from fetal stages to old age (Wegiel et al., 2007), but amyloid deposition, as well as plaque formation, are generally absent in young people. It is also consistent with the absence of plaques when non-secretory Aβ_1–42 is expressed directly in mammalian neuronal cytosol (Jucker et al., 1992; LaFerla et al., 1995) as well as studies showing that both internalization of extracellular Aβ as well as plasma membrane binding are necessary for toxicity in various types of cultured mammalian cells and neurons, and that endocytosis appears to play an essential role in this toxicity (Simakova and Arispe, 2007; Friedrich et al., 2010). Finally, our proposal agrees with the recent observations that autophagy may be necessary for plaque formation in a transgenic mouse model of AD that is dependent on amyloidogenic APP processing (Nilsson et al., 2013).