Presynaptic dystrophic neurites surrounding amyloid plaques are sites of microtubule disruption, BACE1 elevation, and increased Aβ generation in Alzheimer’s disease

E15.5-16.5 C57BL/6 mouse cortical neurons were plated on poly-l-lysine (Sigma) coverslips in 12-well plates (150,000 cells/well) in neurobasal media supplemented with 2 % B-27, 500 μM glutamine, 10 % horse serum, and 2.5 μM glutamate. After 2 h, medium was changed to neurobasal with 2 % B-27, 500 μM glutamine, and 2.5 μM glutamate. After 1–3 days, medium was replaced with neurobasal plus 2 % B-27 and 500 μM glutamine. Following 9 days in culture, neurons were treated with 1 μM Aβ42 oligomers or vehicle prepared as previously described [48, 55]. Briefly, lyophilized recombinant Aβ42 peptide (rPeptide, cat# A-1163-1) was HFIP treated, dried down, then resuspended to 5 mM in dry DMSO (Molecular Probes, # D12345) and brought to 100 μM in cold, 4 mM HEPES pH 8, incubated on ice at 4 °C for 24 h to generate oligomers. For control cultures, DMSO alone was added to 4 mM HEPES pH 8 and incubated as described. After 72 h of Aβ42 treatment, neurons were fixed for 20 min in 4 % paraformaldehyde, 0.12 M sucrose in PBS, permeabilized in 0.5 % Triton, and incubated in anti-βIII tubulin mouse monoclonal antibody (TuJ1, gift of Dr. Lester Binder, 1:500) followed by donkey anti-mouse Alexa 568-conjugated secondary antibody (Molecular Probes, 1:1000) and counterstained with 300 nM DAPI. Coverslips were mounted using Prolong Gold (Molecular Probes), and images acquired on a Nikon A1 confocal microscope with a 60× objective (NA 1.4) and NIS Elements software.

5XFAD mice were generated as previously described [40] and maintained by crossing transgene positive males with B6/SJL F1 hybrid females (Jackson Laboratories). As negative controls, 5XFAD mice were crossed to BACE1^−/− mice [4] to generate 5XFAD;BACE1−/− mice. All animal work was done in accordance with Northwestern University IACUC approval.

Human post-mortem brain tissue was obtained from three AD patients and three cognitively normal controls (Supplementary Table S2) diagnosed at the Cognitive Neurology and Alzheimer’s Disease Center with approval from the Northwestern University IRB. 40-μm floating sections of superior temporal gyrus were stained and imaged as described for murine tissue, except that primary and secondary antibody incubations were 48 h at 4 °C to improve antibody penetration, methoxy XO4 was added after secondary antibody incubation to label plaques, and 0.2 % Sudan Black in 40 % ethanol was used to quench autofluorescence. Sections incubated in parallel without primary antibody were included as negative controls for autofluorescence and background binding of secondary antibody. A minimum of 10 plaques per case were imaged and analyzed on a Nikon A1 confocal microscope with a 60× objective (NA 1.4) and NIS Elements software. Image acquisition settings were maintained the same between AD and control cases. We note that the βIII-tubulin antibody TuJ1 is very well characterized and produced the expected βIII-tubulin staining pattern in cognitively normal controls (not shown) and in normal-appearing neuropil in AD cases (Fig. 3a, c), which served as positive controls. The no-primary antibody negative control produced only weak non-specific background staining in control and AD brain sections (not shown).

Mice were perfused with ice-cold PBS containing protease and phosphatase inhibitors (Calbiochem), brains harvested, and one hemibrain/mouse drop fixed in 4 % paraformaldehyde/PBS and cryopreserved in 30 % w/v sucrose/PBS for sectioning. The other hemibrain/mouse was dissected into cortex and hippocampus and flash frozen in LN2 for biochemistry. 30-μm coronal or sagittal floating brain sections were cut and subjected to antigen retrieval for 1 h at 80 °C in 0.1 M sodium citrate pH 9 followed by 16 mM glycine in Tris-Buffered Saline with 0.25 % triton-X 100 (TBS-T) and blocked in 5 % donkey serum in TBS-T. Sections were incubated overnight at 4 °C with primary antibodies listed in Supplementary Table 1, followed by secondary antibodies (donkey anti-mouse, rabbit or goat conjugated to Alexa 488, 568 or 647; Molecular Probes) used at the same concentration as primary antibody for 2–3 h at room temperature (except 4 °C overnight to improve penetration in the case of anti-tubulin antibodies) plus 300 nM DAPI. Sections incubated in parallel without primary antibody were included as negative controls for autofluorescence and background binding of secondary antibody. Sections were mounted with Prolong Gold (Molecular Probes) and images acquired on a Nikon A1 confocal microscope with a 60× objective (NA 1.4) and NIS Elements software. All image acquisition settings were maintained the same between genotypes (5XFAD;BACE1+/+ and 5XFAD;BACE1−/−). For all proteins whose plaque-associated co-localization with BACE1 was not previously published, 5–8 male and female 5XFAD mice at 5–6 or 9 months of age (Table S3) were analyzed, and 20–40 plaques were imaged for each antibody stain. For co-localization patterns of proteins with BACE1 that have been previously published (Table S4), we did not include additional mice but did confirm the originally reported staining patterns for βIII-tubulin (Fig. 3), MAP2 (Fig. 6), synaptophysin (Fig. 6), APP (Figs. 8, 9), and Aβ/3D6 (Figs. 8, 9). We validated neoepitope antibodies to APP cleavage products using brain sections from 5XFAD;BACE1−/− mice (negative control). These sections were processed and imaged in parallel with brain sections from 5XFAD;BACE1+/+ mice. For other antibodies, we considered normal-appearing neuropil regions distant from plaques to be an internal positive control, as well as normal-appearing regions of the stratum lucidum (Fig. 6, bottom row), which serves as a positive control for normal BACE1 localization and co-localization with other proteins.

5XFAD mouse brain tissue was prepared for electron microscopy as previously described [25]. Briefly, mice were anesthetized with isofluorane (Isothesia, Butler), perfused transcardially with 0.12 M PBS (pH 7.4) for 1 min, then a dilute aldehyde mixture (1 % paraformaldehyde, 1.25 % glutaraldehyde, 0.02 mM CaCl₂ in 0.1 M sodium cacodylate buffer) for 30 min, and a concentrated aldehyde mixture (2 % paraformaldehyde, 2.5 % glutaraldehyde, 0.04 mM CaCl₂ in 0.1 M sodium cacodylate buffer) for 10 min. The brain was removed and placed in ice-cold concentrated fixative on a shaker at 4 °C overnight. The following day brain was bisected, rinsed 3× 20 min in 0.12 M TBS and cut into 70 μm coronal sections. Sections were washed in 0.12 M phosphate buffer (PB) 3× 10 min at 4 °C, treated with 2 % OsO₄ in 0.12 M PB for 1 h at 4 °C, and washed 3× 10 min in 0.12 M PB. The tissue was then dehydrated in graded ethanols and propylene oxide, infiltrated with 1:1 araldite:propylene oxide, flat embedded between aclar sheets and cured for 48 h at 60 °C. Regions of interest were subdissected and re-embedded as above. Serial ultrathin sections (65 nm) were cut, placed onto formvar-coated slotted grids, then stained with 5 % aqueous uranyl acetate for 15 min, followed by Reynold’s lead citrate (1.33 g lead nitrate, 1.76 g sodium citrate, 30 ml distilled water, 8 ml of 1 N NaOH, then diluted to 50 ml) for 10 min, and rinsed in ultrapure dH₂O. Images (7500–20,000×) were acquired on a JEOL 1200EX electron microscope (JEOL Ltd., IL, USA) from 10 to 30 serial sections. ImageJ software was used for processing image stacks of serial sections.

Frozen brain tissue was homogenized in RIPA buffer and 20 μg of homogenate separated by 10 % tris–glycine SDS-PAGE. Protein was transferred onto 0.45 μm PVDF membrane, stained with 0.1 % Ponceau, and imaged. Blots were incubated with anti-APP antibody (6E10, Covance Sig-39300 1:2000) or anti-cathepsin D (Abcam # ab75852 1:1000) followed by HRP-conjugated anti-mouse or anti-rabbit secondary antibody (Vector Laboratories 1:10,000). Blots were visualized using chemiluminescence (Luminata Crescendo, Millipore), signals quantified using a Kodak Image Station 4000 R, normalized to Ponceau, densitometric analyses performed using Kodak 1D 3.6 image analysis software, and statistics analyzed in Prism GraphPad.

Statistical differences for immunoblot experiments were determined using two-tailed Student’s t test (GraphPad Prism Software, Inc., San Diego, CA, USA). Graphed data are presented as the mean ± SEM, and p < 0.05 was considered significant.

In addition to amyloid plaques and neurofibrillary tangles, other hallmark lesions of the Alzheimer’s disease brain include swollen, dystrophic neurites that surround amyloid plaques (reviewed in [9, 51]). Since these dystrophic neurites are predominantly found in very close proximity to, and often in contact with, plaques [25, 62] a neurotoxic form of Aβ (e.g., oligomers) likely causes dystrophy formation. To test this hypothesis, we isolated primary cortical neurons from embryonic mice and exposed them to 1 μM of oligomeric Aβ42 for 72 h, then performed immunofluorescence microscopy for neuron-specific βIII-tubulin. Aβ42-treated neurons displayed processes with numerous βIII-tubulin accumulations along their lengths resembling beads on a string, compared to vehicle-treated neurons (Fig. 1a). Neuritic beading was not likely associated with processes related to cell death caused by high concentrations of Aβ42, since we previously observed that 5 days of treatment with 1 and 2 μM Aβ42 oligomers did not increase caspase 3 cleavages in primary neurons [49]. These beaded structures bore a striking resemblance to tubulin accumulations observed in dystrophic neurites around plaques in vivo (Fig. 3), suggesting a similar mechanism of generation. Our observation of tubulin accumulation in beaded processes in Aβ42-treated cultured neurons is in agreement with extreme beading of primary neurons following 4 days of exposure to 15 μM Aβ42 [11].

Neuritic dystrophy and beading are observed in neurons exposed to a variety of stressors and may be part of a regulated neurite degeneration pathway separate from cellular apoptosis [10, 21, 28]. The presence of abnormal βIII-tubulin accumulations in Aβ42-treated primary neurons suggested that Aβ42 could disrupt microtubule networks, and that impaired axonal transport could be a specific and early effect of Aβ42 leading to serious neuronal and synaptic dysfunction. To test this hypothesis, we followed microtubule dynamics in real time by live-cell imaging to determine if and when Aβ42 could cause microtubule disruption. We incubated primary neurons with Tubulin Tracker (Oregon Green-labeled taxol) that binds only to polymerized tubulin in microtubules, and then exposed the neurons to 10 μM Aβ42 oligomers (Fig. 1b) to model the potentially high Aβ42 oligomer concentrations in the immediate vicinity of amyloid deposits [29]. After 22 h of treatment, neuritic beading that strongly resembled the previously observed aberrant βIII-tubulin immunostaining pattern (Fig. 1a) was apparent in live Aβ42-treated Tubulin Tracker-labeled neuron cultures (Fig. 1b), while vehicle-treated neurons remained healthy with mostly smooth processes. These findings suggested that Aβ42 exposure results in microtubule disruption leading to abnormal tubulin accumulation and neuritic beading in neurons.

To determine how early microtubule disruption and neuritic beading occur after exposure to Aβ42, we used spinning disk confocal microscopy to image live Tubulin Tracker-labeled primary neurons continuously every 5 min after 10 μM Aβ42 oligomer or vehicle treatment. We observed that noticeable microtubule disruption and beading occurred on neurites of primary neurons as early as 3.5 h following Aβ42 treatment, while vehicle-treated cultures appeared normal (Figs. 2a, S1). The very early appearance of disrupted microtubules in beaded neurites suggested that this is one of the primary ways in which Aβ42 oligomers could impair neuronal function. Also, differential interference contrast microscopy indicated that most of the neurites with beads or aberrant microtubule organization were still continuous and intact, although morphologically abnormal (Figs. 2a, S1). These results were highly reproducible, as they were representative of six individual experiments, all performed using separate primary neuron and Aβ42 preparations. We also observed a greater overall decrease in Tubulin Tracker fluorescence after 3 h in the Aβ42-treated neurons compared to vehicle (Fig. 2b), suggesting that Aβ42 causes microtubule depolymerization. To limit photodamage, we performed an identical live-imaging experiment, except capturing an image once every hour for 6 h, and observed even more striking neuritic beading and microtubule disruption in Aβ42-treated neurons, while vehicle-treated neurons showed very little change (Fig. 2c).

Aβ42 oligomer concentrations in the range of 1–10 μM induce cell death in primary neuron culture (reviewed in [16]). To determine whether neuritic beading and microtubule disruption were the results of Aβ42-induced cell death processes, we repeated Tubulin Tracker labeling of primary neurons followed by Aβ42 treatment for 3.5 h, then added propidium iodide to label the nuclei of dead cells (Fig. S1). Although we noted Aβ42-induced neuritic beading and microtubule disruption (Fig. S1a), the percentage of live cells as measured by absence of propidium iodide uptake was identical for both Aβ42-treated and vehicle control neurons (Fig. S1b). This result demonstrates that neuritic beading and microtubule disruption precede the induction of cell death in Aβ42-treated primary neurons.

The Aβ42-induced reduction and aberrant spatial distribution of microtubules in neurites suggested that microtubule-based transport would be impaired in Aβ42-treated primary neurons. It has also been reported that trafficking of BDNF [44] mitochondria [70], and other axonal cargo [57] is impaired by Aβ, so we hypothesized BACE1 transport may be affected as well. To test this hypothesis, we transfected primary neurons with BACE1-YFP and neuropeptide Y-mCherry (NPY-mCherry) fusion constructs, treated with 10 μM Aβ42 oligomers or vehicle, and performed live-cell imaging and kymograph analysis to assess vesicular movement in neurites. We observed that the proportion of motile BACE1-YFP and NPY-mCherry puncta was decreased in neurites of Aβ42-treated neurons, compared to vehicle (Fig. S2). We infer that the effect of Aβ42 on trafficking in neurites is not limited to BACE1-containing vesicles, as the movement of NPY-mCherry puncta was also decreased. Taken together, our results suggest that Aβ42 causes microtubule depolymerization and network disruption, leading to neuritic beading and impaired microtubule-based axonal transport in neurons.

In both AD and APP transgenic mouse brains, BACE1 accumulates in swollen presynaptic dystrophic neurites that surround amyloid plaques in close proximity [25, 85, 86]. To gain insight into the mechanism of Aβ-induced BACE1 elevation in peri-plaque dystrophies and the potential role of microtubules, we performed immunofluorescence staining of superior temporal gyrus from AD patients and cognitively normal controls (Table S2) with antibodies against BACE1 and neuron-specific βIII-tubulin, as well as methoxy XO4 to label fibrillar amyloid deposits. As we have previously reported [25, 86], in AD brain BACE1 immunoreactivity was elevated in dystrophic neurites in a halo pattern immediately surrounding the methoxy XO4-positive amyloid plaque core. Interestingly, we observed a striking reduction of βIII-tubulin signal within the BACE1-positive peri-plaque halo, while the pattern of βIII-tubulin immunoreactivity seemed unaffected in nearby normal-appearing neuropil (Fig. 3a, first two columns; Fig. S3, first column) and in cognitively normal controls (not shown). To determine the quantitative relationships between BACE1 and βIII-tubulin immunoreactivities within the peri-plaque halo, we calculated the ratio of BACE1 to βIII-tubulin immunofluorescence intensities in dystrophic neurites (defined by high BACE1 signal) around plaques and in regions of normal neuropil distant from plaques for the AD cases (Fig. 3b). Importantly, we observed that the peri-plaque BACE1:βIII-tubulin ratio was significantly elevated (Fig. 3b, first two panels). We also generated intensity profiles of BACE1 and βIII-tubulin immunofluorescence signals through the plaque, and generally found an inverse relationship between BACE1 and βIII-tubulin intensities, such that high BACE1 signal typically occurred in areas of low βIII-tubulin intensity (Fig. 3c, first two columns). In cognitively normal controls, no plaques, and hence no areas of elevated BACE1 or loss of βIII-tubulin, were observed (not shown). These results confirm and extend our initial observation of decreased βIII-tubulin staining around plaques in AD [86]. Moreover, they suggest that reduced tubulin level, and hence low microtubule density, in dystrophic regions surrounding plaques is a pathologic feature of AD that is the consequence of short-range Aβ toxicity to neurites, and may have a role in BACE1 accumulation.

Post-mortem AD brain samples represent end-stage disease and thus provide limited information about pathogenesis. To gain insight into the development of Aβ-induced dystrophic neurites, we turned to our 5XFAD mouse model of amyloid pathology [40]. Our previous work showed that brains of AD patients and 5XFAD mice have elevated levels of BACE1 that accumulate in dystrophic presynaptic neuronal structures (likely axons and terminals) in a halo that surrounds the amyloid deposit [86]. Taken together, these and other data suggest that the 5XFAD mouse is a faithful model of BACE1 elevation in AD.

To assess the distributions of tubulin and BACE1 around amyloid deposits in our AD mouse model, we immunostained brain sections from five different 5- to 6-month-old 5XFAD mice with antibodies directed against specific isoforms of tubulin, including neuron-specific βIII-tubulin, acetylated α-tubulin, and polyglutamylated tubulin (Fig. 3a, third, fourth and fifth columns, respectively; Figs. S3, S4). BACE1 immunoreactivity was concentrated in a halo surrounding the core of the amyloid deposit, as previously reported [25, 85, 86]. Interestingly, similar to human AD, we observed that βIII-tubulin immunostaining was largely absent in BACE1-positive peri-plaque dystrophies. The βIII-tubulin signals that did occur in the peri-plaque halo appeared as amorphous accumulations with a staining pattern complementary to that of BACE1, i.e., regions of high βIII-tubulin intensity generally showed low BACE1 signal and vice versa. Acetylated and polyglutamylated isoforms of tubulin showed similarly reduced patterns of immunostaining in the peri-plaque halo. Additionally, we measured peri-plaque BACE1:tubulin ratios and intensity profiles for 5XFAD amyloid deposits (Fig. 3b, c) and found increased ratios and inverse relationships between BACE1 and tubulin isoforms, similar to our results for human AD plaques. We note that the peri-plaque human and mouse tubulin and BACE1 immunoreactivity patterns are similar but not identical, probably because mouse deposits are at most a few months old while human plaques are possibly many years old. At the end-stage of disease in post-mortem AD brain, there are very few, if any neurites near plaques that are not severely degenerated, thus potentially explaining the lower intensities of BACE1 and tubulin immunostaining in peri-plaque dystrophies of human AD compared to 5XFAD mouse.

The reduction in tubulin isoforms in 5XFAD BACE1-positive dystrophies suggested that microtubules are decreased, disorganized, or disrupted in presynaptic dystrophic neurites near amyloid deposits. To investigate this possibility at an ultrastructural level, we examined ultrathin sections of 5XFAD mouse brain by electron microscopy (EM). We found swollen dystrophic neurites in abundance in close proximity or virtually contacting amyloid deposits (Fig. 4a). We confirmed the close physical association between BACE1-positive dystrophies and amyloid deposits in 3D reconstructions of multi-photon confocal microscopy images of live brain slices of BACE1-YFP;5XFAD multi-transgenic mice (Supplementary Text; Fig. S5, Videos S1, S2). By BACE1 immuno-EM, we previously reported that dystrophies surrounding plaques are engorged with large BACE1-negative electron-dense multi-lamellar vesicles and smaller BACE1-positive electron-translucent vesicles [25]. Here, by EM we found dystrophies that were definitively identified as axons, as indicated by the presence of myelin sheaths (Fig. 4b–f). Importantly, we found occasional EM series displaying normal-appearing axons with dense microtubule bundles that enter a dystrophic region within which microtubules are strikingly absent, reduced in length, or dramatically disorganized (Fig. 4c–f). These data suggest that in human AD and in amyloid pathology mouse models, peri-plaque dystrophic axons are deficient in normal functional microtubule networks for axonal transport. Vesicles may enter but cannot transit or exit dystrophic regions due to absence of microtubules, which could explain the dramatic accumulation of BACE1 around plaques.

Recent work has linked mutations in various proteins of the dynein–dynactin complex required for retrograde axonal transport to neurodegenerative disease in humans and mouse models [43]. If microtubule networks are disrupted in dystrophies around plaques, we reasoned that the microtubule motor proteins that move along them carrying cargo from the terminals back to the cell body could also be mis-localized, compounding the impaired removal of vesicles from dystrophic axons and terminals. This could lead to decreased turnover of proteins that are preferentially localized to presynaptic terminals, like BACE1 [25, 86]. To determine the localization pattern of microtubule motor proteins in peri-plaque dystrophies, we performed immunofluorescence microscopy of coronal brain sections from five individual 5- to 6-month-old 5XFAD mice using antibodies directed against cytoplasmic dynein intermediate chain (DIC), two subunits of dynactin (p150glued and dynamitin), and kinesin heavy chain (Figs. 5, S3, S4). DIC is a component of the retrograde microtubule motor, and while we observed some DIC staining in BACE1-positive dystrophic neurites, it was quite low compared to the amount of potential cargos such as BACE1 that accumulate in dystrophies (Figs. 5a, S3, S4). p150glued is required to initiate retrograde transport from the axon terminal [32, 37], stabilize microtubules in axons [30], and improve processivity of dynein [27, 35]. Importantly, p150glued was diminished in BACE1-positive peri-plaque dystrophies compared to normal surrounding neuropil (Figs. 5b, S3, S4), suggesting a deficiency of functional and active retrograde dynein–dynactin complexes in these dystrophies. In contrast, dynamitin staining was noticeably increased in some, but not all, BACE1-positive peri-plaque dystrophies (Figs. 5c, S3, S4). Dynamitin tethers the cargo-binding domain of dynactin to the microtubule and dynein-binding domain through its interactions with Arp1 and p150glued, respectively [5]. The increased dynamitin in peri-plaque dystrophies suggested that it may be trapped in either free form or in complexes with adaptor proteins and cargo, but lacking p150glued, dynamitin cannot be attached to dynein and transported with cargo back toward the soma. Disruption of microtubule networks in peri-plaque dystrophic neurites would also be predicted to have an effect on anterograde axonal transport by kinesins and cause kinesin mis-localization. Our immunofluorescence microscopy analysis showed that, similar to dynamitin, kinesin heavy chain accumulated in dystrophies surrounding plaques (Figs. 5d, S3, S4), supporting our hypothesis that peri-plaque presynaptic dystrophic neurites are deficient in microtubules and exhibit mis-localized microtubule motor proteins, conditions that would impair axonal transport.

To further characterize BACE1-positive peri-plaque dystrophic neurites in 5XFAD brain sections, we performed immunofluorescence microscopy with antibodies directed against BACE1 and several other neuronal proteins (Figs. 6, S3, S4), confirming that accumulation in dystrophies is specific to certain proteins but not others. Synaptophysin and bassoon, both presynaptic proteins, show strong co-localization with BACE1 in the stratum lucidum, the axon terminals of the mossy fiber pathway in the hippocampus (Fig. 6a, b; bottom row), confirming our previous results that BACE1 is normally localized in the presynaptic terminal [25]. As we have reported before [25, 86], and was recently confirmed [15, 50], many peri-plaque dystrophies appear to be axonal in origin, showing accumulations of the presynaptic protein synaptophysin (Fig. 6b), and a conspicuous absence of the somatodendritic marker MAP2 (Fig. 6c), which does not co-localize with BACE1 in the stratum lucidum (Fig. 6c; bottom panel). Surprisingly, given its strong co-staining with BACE1 in normal tissue of the stratum lucidum, bassoon appears to be entirely absent from peri-plaque dystrophies (Fig. 6a). Since bassoon is a well-established marker of synaptic active zones (reviewed in [6]), its absence from peri-plaque dystrophies indicates a lack of functional synapses in these regions. The toxic effect of amyloid plaques on synapses seems to be quite local, as the immunofluorescence signal for bassoon outside the immediate peri-plaque halo of the dystrophies appears as a normal punctate pattern. This is supported by other APP transgenic mouse studies showing that PSD95 was reduced by 60 % in the halo of oligomeric Aβ42 surrounding plaques, while 50 μm from the plaque it was found at normal levels [29].

Cathepsin D is found in lysosomes, which normally localize to a peri-nuclear region of the soma rather than axon terminals, and cathepsin D does not co-localize with BACE1 in normal-appearing regions of the 5XFAD stratum lucidum (Fig. 6d, bottom panel). Although cathepsin D appeared somewhat elevated in 5XFAD BACE1-positive peri-plaque dystrophies, the patterns of cathepsin D and BACE1 localization did not overlap as extensively as those of synaptophysin and BACE1 (Figs. 6d, S3, S4). BACE1 is normally degraded through lysosomal pathways [26, 60], and thus presynaptic BACE1 is dependent on retrograde transport to the cell body for turnover. Early lysosomes, identified by Lamp1 expression, are found in dystrophic neurites [15, 25] but are deficient in cathepsins [15]. These studies together with our results suggest that microtubule disruption and motor protein mis-localization impair retrograde transport and hence lysosomal maturation, ultimately resulting in reduced degradation of BACE1 and other proteins in peri-plaque dystrophic neurites. The cathepsin D antibody that we used for our immunolocalization analysis (Fig. 6d) recognizes the C-terminus of both 43 kDa pro-cathepsin D and 28 kDa cathepsin D heavy chain, precluding our ability to determine whether peri-plaque cathepsin D was mature (active) or immature (inactive). To distinguish and quantify mature and immature forms of cathepsin D, we performed immunoblot analysis of brain homogenates from 6-month-old 5XFAD and non-transgenic mice using the cathepsin D antibody (Fig. 7a). We observed that 5XFAD brains had significantly increased levels of immature pro-cathepsin D compared to brains of non-transgenic littermates, while 5XFAD levels of mature cathepsin D heavy chain were unchanged (Fig. 7b). This result is consistent with the notion of impaired maturation of lysosomes in the brains of 5XFAD mice, which might contribute to accumulation of BACE1 and other proteins in peri-plaque dystrophic neurites.

Our observations of BACE1 accumulation around plaques led us to the hypothesis that elevated levels of BACE1, APP [25, 86], and PS1 [81] found in peri-plaque dystrophic neurites would cause increased BACE1 processing of APP and Aβ production, thus exacerbating AD pathogenesis. However, direct evidence that elevated BACE1 cleaves APP and leads to increased Aβ formation in dystrophic neurites has been lacking. To address this question, we made use of several antibodies specific to neoepitopes of APP-derived fragments revealed by BACE1 and γ-secretase processing. BACE1 cleavage of APP generates two products, the N-terminal soluble APP ectodomain fragment, sAPPβ, and the membrane bound C-terminal fragment, C99. The antibody ANJJ [45] recognizes the free C-terminus of sAPPβ created by BACE1 processing of APP with the Swedish familial AD mutation [38], which is expressed in 5XFAD mice [40]. To validate ANJJ for tissue immunostaining and verify that it does not recognize unprocessed full-length APP, we co-stained sagittal brain sections from 9-month-old 5XFAD;BACE1+/+ and 5XFAD;BACE1−/− bigenic mice with ANJJ and an antibody to an N-terminal epitope of full-length APP [64] and performed immunofluorescence confocal microscopy (Fig. 8). We observed punctate sAPPβ immunoreactivity of ANJJ in the soma and processes of 5XFAD;BACE1+/+ large pyramidal neurons in the cortex (Figs. 8a, upper panels; S4), the primary cells that express the 5XFAD transgenes in the brain [40]. The punctate intracellular immunostaining pattern for sAPPβ was expected, since BACE1 cleaves APP in endosomes [45, 66] to create the sAPPβ neoepitope and release the soluble APP ectodomain into the endosome lumen for eventual secretion into the extracellular milieu. The immunoreactivity of full-length APP exhibited a wider spatial distribution in the neuron compared to that of sAPPβ, indicating cell surface localization in addition to intracellular labeling. In contrast, sAPPβ immunosignal was completely absent in 5XFAD;BACE1−/− brain sections (Fig. 8a, lower panels), indicating the absence of the sAPPβ free C-terminal neoepitope as expected for BACE1-deficient mice [34], even though full-length APP immunoreactivity throughout the neuron was similar to that observed in 5XFAD;BACE1+/+ sections. These results demonstrate conclusively that the ANJJ antibody only recognizes the BACE1-cleaved sAPPβ C-terminal neoepitope and does not cross-react to full-length APP in brain sections.

We also performed immunofluorescence microscopy on 5XFAD;BACE1+/+ and 5XFAD;BACE1−/− brain sections incubated with antibodies recognizing neoepitopes on the free N-terminus of BACE1-cleaved Aβ (3D6) [23] and the free C-terminus of γ-secretase-cleaved Aβ ending in amino acid 42 (Aβ42), co-stained with the full-length APP antibody (Fig. 8b). As with sAPPβ immunosignal, we observed intraneuronal 3D6 and Aβ42 neoepitope immunostaining only in large pyramidal cortical neurons of 5XFAD;BACE1+/+ mice, while those of 5XFAD;BACE1−/− mice were completely devoid of either 3D6 or Aβ42 neoepitope immunoreactivities (Fig. 8b) even though they showed robust full-length APP labeling. Taken together, these results fully validate the 3D6 and Aβ42 neoepitope antibodies for tissue immunostaining and confirm that they do not label full-length APP.

Next, using our validated neoepitope antibodies we sought to determine whether elevated BACE1 in peri-plaque dystrophies was associated with increased BACE1 processing of APP and Aβ generation. Importantly, sagittal brain sections from 5- to 6-month and 9-month-old 5XFAD mice co-incubated with sAPPβ neoepitope and BACE1 antibodies showed robust punctate sAPPβ immunoreactivity, likely representing endosomal localization, within peri-plaque dystrophies that significantly co-localized with BACE1 immunosignal (Figs. 9a, S4). sAPPβ immunoreactivity was very intense in peri-plaque dystrophies, compared to the very low sAPPβ immunosignal in the surrounding neuropil, indicating that BACE1 processing of APP was dramatically increased in the BACE1-positive dystrophic neurites near amyloid deposits. We noted that peri-plaque dystrophies exhibited variable levels of transgenic sAPPβ immunoreactivity, presumably because the Thy1 transgene promoter is expressed in some, but not all, cortical neurons [36]. Similar to the sAPPβ immunostaining pattern, we observed increased punctate 3D6 and Aβ42 neoepitope immunosignals that co-localized significantly with BACE1 immunoreactivity in peri-plaque dystrophies (Figs. 9b, c, respectively; S4), suggesting that BACE1 accumulation elevated Aβ production. Although 3D6 and Aβ42 immunosignal intensities in dystrophies were weaker than those in amyloid deposits, they were obviously elevated compared to surrounding neuropil, especially at higher magnifications (right panels, 9b, c). Additionally, sAPPβ and 3D6 neoepitope immunoreactivities co-localized significantly in peri-plaque dystrophies (Fig. 9d). Co-immunostaining of 5XFAD brain sections with full-length APP and BACE1 antibodies confirmed that APP and BACE1 accumulate together in peri-plaque dystrophies (Fig. 9e), implicating both in plaque progression.

Taken together, our results suggest that elevated BACE1 in dystrophic neurites around plaques is associated with increased BACE1 cleavage of APP and Aβ generation. The presence of neoepitope N-terminal and C-terminal APP processing products in peri-plaque dystrophies strongly supports the hypothesis that BACE1 and γ-secretase are active in dystrophies, although we cannot formally exclude the possibility that APP processing occurs elsewhere in the cell followed by APP product trafficking and accumulation in dystrophies. Additionally, our results show that peri-plaque dystrophic neurites accumulate Aβ42, which if released into the extracellular space could contribute to plaque growth and further neuritic dystrophy.