Introduction
Synaptic dysfunction is an early event in Alzheimer’s disease (AD). Soluble oligomers of amyloid β (Aβ), which are generated from the amyloid precursor protein (APP), are believed to be the primary synaptotoxins in AD. According to one central view of AD pathogenesis, extracellular Aβ oligomers (eAβOs) bind plasma membrane targets to elicit pre- and postsynaptic intracellular effects (for reviews, [
1,
2]). At the postsynaptic membrane, eAβOs interact with glutamate receptors and dysregulate calcium influx to impair long-term potentiation (LTP) and enhance long-term depression (LTD) [
3‐
6]. eAβO binding also alters spine morphology and decreases spine density [
7,
8]. In axons, eAβOs impair transport of cargoes such as mitochondria and vesicles containing brain-derived neurotrophic factor (BDNF) [
9‐
11], which are both required for neuronal form and function. Mitochondria are needed at presynaptic boutons to maintain neurotransmission by producing ATP and buffering synaptic calcium (Ca
2+) [
12‐
14]. Once secreted from axon terminals, BDNF increases spine density and the proportion of mature spines by interacting with postsynaptic TrkB receptors at the target cell membrane [
15,
16]. Thus, impaired transport of mitochondria and BDNF might contribute to synaptic dysfunction in AD [
17,
18].
Although a role for eAβOs in causing AD-like toxicity is well established, several studies have revealed that intraneuronal accumulation of Aβ is also toxic and precedes its extracellular deposition in patients and model mice of AD [
19‐
24]. In model mice that overexpress mutant human APP, synaptic dysfunction [
22,
25], spine morphology alteration [
26], and axonal transport defects [
27] were observed in association with intracellular Aβ oligomers (iAβOs). These pathological changes, however, may have been induced by overexpression of human APP or undetected extracellular Aβ. Cellular mechanisms that underlie iAβO-induced synaptic dysfunction remain uncharacterized. Furthermore, although it is widely reported that tau is required for eAβO toxicity (for a review, [
28]), whether iAβO toxicity is tau-dependent has not yet been investigated.
A valuable model for studying iAβOs is an APP mutation identified in familial AD. The Osaka (E693Δ) mutation in APP induces iAβO accumulation without detection of Aβ fibrillization
in vitro and without detectable Aβ plaque formation in AD patients or mouse models [
29‐
31]. Intracellularly, the Osaka mutation-induced iAβOs lead to endoplasmic reticulum stress and damage of mitochondria and organelles within the endosomal/lysosomal system [
32].
Here, we determined and compared the effects of wild-type APP (APPWT) and Osaka-mutant APP (APPOSK) on dendritic spine morphology and intracellular transport of organelles required for synaptic maintenance and function. We found that iAβOs reduced the number of mature spines and impaired transport of BDNF, mitochondria, and recycling endosomes in hippocampal neurons expressing APPOSK. Notably, spine reduction and impairment of BDNF transport occurs independently of tau. These results advance our understanding of early AD synaptic pathology because iAβO accumulation precedes extracellular amyloid deposition in patients and AD model mice. Our findings may promote development of effective therapeutic compounds for AD prevention and treatment.
Materials and methods
Preparation of primary neurons
Mouse primary neurons were prepared from embryos of wild-type (
MAPT +/+) and tau knockout (
MAPT −/−) mice (Jackson Laboratory, Bar Harbor, ME) at embryonic day 18 (E18). Hippocampal tissues were dissected in ice-cold Hank’s balanced salt solution (HBSS; Sigma-Aldrich, St. Louis, MO) and incubated in 1 ml of papain solution (2 mg/ml in HBSS) at 37 °C for 30 min with gentle mixing. After being washed with 5 ml of 50 % horse serum in HBSS once, 5 ml of HBSS twice, and 4 ml of neuronal culture medium (Neurobasal, Electro medium supplemented with B27, Electro and 500 μM L-glutamine; all from GIBCO, Life Technologies, Carlsbad, CA) once, the tissues were dissociated into cells by pipetting several times with a Pasteur pipet in 2 ml of neuronal culture medium. The cell suspensions were plated onto poly-L-lysine-coated coverslips in 6-well culture plates at a density of 170,000 cells/2 ml/well. For transport analyses, primary neurons from wild-type and tau knockout mice, and wild-type rats were prepared from embryos at E16 and E18, respectively, as described [
33]. Neurons were cultured on coverslips in Neurobasal medium supplemented with B27 at a density of 250,000 cells/5 ml in 6 cm dishes. The astrocyte feeder layer for the neuronal co-culture was generated using neural progenitor cells as described previously [
34].
Expression vectors
The pCI-APP construct was prepared using a pCI vector (Promega, Madison, WI) as described previously [
29]. A pEGFP (enhanced green fluorescent protein)-N2 vector was obtained from Clontech (Takara Bio Inc. Otsu, Japan). The pIRES2-APP-EGFP vectors were made by amplifying APP
WT and APP
OSK using a forward primer containing an Nhe I restriction site (underlined) 5′-AATTAATTAA
GCTAGCGCCACCATGGGGGCTGCCCGGTTTGGCACTGCT-3′ and a reverse primer containing a Sac II restriction site 5′- TTAATTAATT
CCGCGGCTAGTTCTGCATCTGCTCAAAGAACTTGTAGGTTGG-3′ for subcloning into pIRES2-EGFP (Takara Clontech Bio Inc. Otsu, Japan). Plasmid composition was confirmed by sequencing. The pβ-actin-BDNF-mRFP (monomeric red fluorescent protein) and pβ-actin-eBFP (enhanced blue fluorescent protein) vectors were a kind gift from Dr. G. Banker (Oregon Health and Sciences University). The pcDNA3 Mt-eYFP (enhanced yellow fluorescent protein) construct was designed to express the COX IV mitochondrial-targeting sequence-EYFP fusion proteins and kindly gifted from Dr. G. Rintoul (Simon Fraser University). The JPA5-TfR -GFP vector is described in Burack et al. [
35].
Analyses of APP expression
Mouse primary neurons were cultured for 21 days in vitro. The cells were cotransfected with pCI-APP (APPWT, APPOSK, or empty) and pEGFP-N2 using a Lipofectamine2000 reagent (Invitrogen, Life Technologies). Transfection was performed in the presence of 0.5 μM kynurenic acid (Sigma-Aldrich) to lessen excitotoxic cell damage. Cells expressed the transgenes for 2 days. For immunocytochemical analysis of APP expression, the cells were fixed with 4 % paraformaldehyde in PBS at room temperature for 30 min and permeabilized by immersion in 0.05 % Tween-20 in PBS for a moment. After a brief wash, the cells were blocked with 20 % calf serum in PBS at room temperature for 1 h. The cells were then stained with human APP-specific antibody 6E10 (Covance, Berkeley, CA) or Aβ oligomer–specific antibody 11A1 (IBL, Fujioka, Japan) at room temperature for 1 h followed by Rhodamine-conjugated anti-mouse IgG antibody (Jackson ImmunoResearch Labs, West Grove, PA) at room temperature for 20 min. The stained specimens were mounted with VECTASHIELD mounting medium with DAPI (H-1500; Vector Laboratories, Burlingame, CA) and viewed under a Leica TCS SP5 confocal laser microscope (Leica, Wetzlar, germany). For Western blot analysis of APP expression, the cells were lysed in 1 % Triton X-100/Tris-buffered saline (100 mM Tris–HCl, pH 7.6, 150 mM NaCl) containing protease inhibitor cocktail P8340 (Sigma-Aldrich). The lysates were separated by SDS-PAGE and transferred onto polyvinylidene difluoride membranes. Human APP and actin were stained with 6E10 and rabbit anti-actin antibody (Sigma-Aldrich) followed by horseradish peroxidase–conjugated secondary antibodies and chemiluminescent peroxidase substrate (Millipore, Billerica, MA). Signals were visualized and quantified using a LAS-3000 luminescent image analyzer (Fujifilm, Tokyo, Japan).
Analyses of dendritic spines
Mouse primary neurons transfected with pCI-APP and pEGFP-N2 were fixed with 4 % paraformaldehyde after a 2-day culture. The fixed cells were mounted, and the images were taken using a Leica TCS SP5 confocal laser microscope. Dendritic spines were classified into four groups by the criteria as follows: mushroom, the length ≤ 5 μm and the ratio of neck width/head width ≥ 1.5; stubby, the length ≤ 1 μm and the ratio of neck width/head width < 1.5; thin, 1 < the length ≤ 5 μm and the ratio of neck width/head width < 1.5; and fillopodia, the length ≥ 1.5 μm without a head [
36,
37]. Three to seven independent cultures were made for each APP
WT-, APP
OSK-, and mock transfection. One transfected cell was chosen from each culture and analyzed for spines with 3 to 12 dendrites per cell.
To study the effects of extracellular Aβ on spines, we first determined the levels of Aβ secreted from cells into culture media. Culture media of mouse primary neurons transfected with pCI-APP were harvested 2 days after transfection. Aβ concentrations in the media were measured using a human/mouse Aβ40 ELISA kit (Wako Pure Chemical Industries, Osaka, Japan). Then, mouse primary neurons transfected with pEGFP-N2 alone (without pCI-APP) were cultured for 2 days in the presence of synthetic wild-type Aβ42 or Osaka (E22Δ)-mutant Aβ42 (41 amino acids) peptide (both from Peptide Institute, Mino, Japan) at various concentrations. After cell fixation, dendritic spines were analyzed as described above.
Analyses of axonal and dendritic transport
Mouse and rat primary neurons were cultured for 10–12 days
in vitro. For analysis of BDNF transport, mouse primary neurons were doubly transfected with pIRES2-APP-EGFP and pβ-actin-BDNF-mRFP. For analysis of mitochondria transport, rat primary neurons were triply transfected with pCI-APP, pcDNA3 Mt-eYFP, and pmUBa-eBFP. To analyze recycling endosome transport, we chose the transferrin receptor (TfR) as a marker for recycling endosomes and used pCI-APP, JPA5-TfR-GFP, and pmUBa-eBFP in transfection of rat primary neurons. The soluble GFP and BFP are distributed throughout the cell body and processes enabling us to determine the orientation of the cell body relative to the axon, thus, distinguishing anterograde and retrograde transport. All transfections were done in the presence of 0.5 μM kynurenic acid as described above. Two days after transfection, BDNF/mitochondria/TfR transport was analyzed using a standard wide-field fluorescence microscope equipped with a cooled charge-coupled device camera and controlled by MetaMorph (Molecular Devices, Sunnyvale, CA) according to Kwinter et al. [
38]. BDNF imaging was recorded by the “stream acquisition module”, and mitochondria and TfR imaging were recorded by the “acquire timelapse module” in MetaMorph. Briefly, cells were sealed in a heated imaging chamber, and streaming recordings of BDNF were acquired from triple transfectants for 25 s (250-msec exposures). Frames were captured continuously for 300 s (400-msec exposures) for mitochondria transport and 100 s (500-msec exposures) for TfR transport. This captured dozens of transport events per cell in 100-μm segments of the axon and 45-μm segments of the dendrite. Axons and dendrites were initially identified based on morphology and confirmed retrospectively by immunostaining MAP2, a dendrite-specific microtubule-associated protein, with mouse anti-MAP2 antibody (Millipore). Vesicle flux was obtained through tracing kymographs in MetaMorph. Vesicle flux was defined as the total distance traveled by vesicles standardized by the length and duration of each movie (in micron-minutes): ∑
i = 1
n
d
i
/(
ℓ ×
t), where
d is the individual vesicle run lengths,
ℓ is the length of axon or dendrite observed, and
t is the duration of the observation [
38].
Statistical analyses
All values obtained are expressed as the mean ± SEM. Comparisons of means among multiple groups were performed using Fisher’s PLSD test following ANOVA. Differences with a p value of <0.05 were considered significant.
Discussion
In the present study we showed that iAβOs, generated by APP
OSK expression in cultured neurons, impaired axonal and dendritic transport of BDNF, mitochondria, and dendritic recycling endosomes (Figs.
5,
7 and
8). Because these cargoes are critical for spine formation and maintenance, reduced trafficking may have led to the observed decreases in spine density and in the number of mature mushroom spines (Fig.
2). Our results are significant because intraneuronal accumulation of Aβ precedes its extracellular deposition in patients and AD model mice; thus, iAβOs likely contribute to the early synaptic pathology in AD. Moreover, multiple lines of evidence demonstrate that trafficking defects are either an early cellular pathology or even causal in AD [
52], underscoring the importance of defining iAβO mechanisms of action.
BDNF transport defects in APP
OSK neurons may reduce the amount of BDNF available for secretion, and in turn, compromise dendritic spine maturation and density. Spines are the primary site of excitatory input on neurons, and a reduced spine number and changes in morphology contribute to synaptic pathology in AD. BDNF secreted from cells binds to and activates TrkB receptors that are located on both presynaptic axon terminals and postsynaptic dendritic spines of glutamatergic synapses [
53]. BDNF-induced TrkB signaling modulates synaptic transmission by enhancing presynaptic glutamate release and increasing the open probability of postsynaptic NMDA receptor ion channel [
53]. Increased NMDA receptor currents activate the Rac1 (Ras-related C3 botulinum toxin substrate 1) pathway and suppress cofilin, an actin-depolymerizing factor, thereby promoting spine growth and stabilization [
54]. Thus, iAβO-induced transport impairment may reduce the amount of BDNF available for secretion, leading to synaptic impairment and spine reduction.
We also found that iAβOs impair transport of mitochondria and recycling endosomes, which are essential for spine development and maintenance. Mitochondria translocate to pre- and postsynaptic regions to supply ATP for neurotransmission [
13]. A reduction in ATP availability may reduce presynaptic secretion of essential signaling molecules such as glutamate and BDNF, and thereby, as similarly described above, impair postsynaptic signaling of cascades such as Ca
2+/calmodulin-dependent protein kinase II (CaMKII) and Rac 1 that are required for spine maintenance [
54‐
56]. A role for dendritic mitochondria is to buffer synaptic calcium; excess calcium may negatively regulate actin-binding proteins that are required for maintaining for spine density and plasticity (discussed below) [
12,
57]. Thus, perturbations in mitochondrial motility or morphology induced by iAβOs likely diminish spine structure stemming from multiple mechanisms. Recycling endosomes also play an essential role in spine morphology. Elegant studies from the Ehlers laboratory demonstrated that active transport of recycling endosomes supplies necessary plasma membrane lipids and proteins required to support spine structure and function [
51]. Taken together, iAβO-induced intracellular transport disruption is likely an important contributor to mature spine loss in APP
OSK-expressing neurons.
iAβOs may disrupt intracellular transport by perturbing motor function through second message cascades or by directly binding to motor proteins. eAβO binding to membrane targets results in changes in kinase and phosphatase activity leading to transport blockade of organelles, including mitochondria and BDNF-containing vesicles [
9‐
11,
48,
58‐
60]. Similarly, organelle transport is reduced in squid axoplasm perfused with AβOs via activation of casein kinase 2 [
61], providing evidence that iAβOs impinge on signaling cascades that reduce trafficking. It is likely that APP
OSK dysregulates intracellular Ca
2+ signaling via ER stress and mitochondrial damage [
32] that may ultimately have negative consequences on the mechanisms that regulate transport. For example, mitochondria transport is governed by Miro-Milton-Kinesin-I in a Ca
2+-dependent mechanism, where Ca
2+ binding to Miro inhibits Kinesin-I-based motility [
62]. It is thus possible that the APP
OSK effects on intracellular Ca
2+ result in a blockade of mitochondria transport. Moreover, Ca
2+ signaling may represent a general mechanism for the regulation of microtubule-based transport. The transport of dense core vesicles containing neuropeptides [
63] and the dendritic kinesin, KIF17, ferrying the NR2B glutamate receptor subunit [
64], are also subject to Ca
2+ regulation, demonstrating that perturbations in Ca
2+ homeostasis likely have broader effects on organelle trafficking. A second possible mechanism of transport disruption is the direct binding of iAβOs or Aβ to motor proteins [
65]. Ari et al. demonstrate mislocalization of NMDA receptors and NGF/NTR (p75) at the post-synaptic membrane due to intracellular Aβ binding the mitotic kinesin Eg 5 [
66].
Despite a plethora of recent reports, the role of tau in transport disruption is still a matter of debate. In vivo studies demonstrate that axonal transport is unaffected by tau overexpression or suppression, or by moderate amounts of hyperphosphorylated tau [
67,
68]. In primary culture, we have demonstrated that eAβO-induced BDNF-transport disruption is independent of tau [
48,
49]. However, other studies suggest that phospho-tau inhibits fast axonal transport (FAT) by interacting directly with motor-cargo complexes or initiating aberrant signaling cascades that alter FAT dynamics [
69,
70], and that tau reduction prevents AβO-induced defects in mitochondria and neurotrophin receptor TrkA transport [
10,
60]. It is possible that transport defects are motor and/or cargo dependent. For example, mitochondria are transported primarily by KIF5 whereas BDNF is transported primarily by KIF1A [
71]. Kinesins may be differentially affected by hyperphosphorylated tau, thus, we cannot rule out that in our studies the reduction in mitochondria and endosome transport is tau-dependent. How APP
OSK and tau influences organelle transport is a focus of ongoing studies.
Similar to the initiation of transport deficits, iAβO-induced signaling cascades may compromise dendritic spine maturation and density. Actin, a critical structural component of spines, is tightly regulated by actin-binding proteins and their associated kinases and phosphatases [
72]. eAβOs reduce spine density and alter morphology by at least two mechanisms: one is the mislocalization of p21-activated kinase (PAK), and another is the activation of the calcineurin, a calcium-dependent phosphatase implicated in AD. Changes in these signaling cascades alter the dynamics of actin-binding proteins needed to maintain and stabilize spine actin [
73‐
75]. Through a yet unexplored mechanism, iAβOs may also impinge on the regulation of actin-binding proteins that lead to spine retraction. Consistent with our previous findings for eAβOs, we show that iAβO-induced BDNF transport defects and spine loss occur independent of tau. These effects are likely caused by signaling cascade activation, such as the calcineurin-GSK3β pathway, that persists in the absence of tau [
48]. Because tau knockout mice do not exhibit any serious health or cognitive deficits, some researchers proposed that lowering endogenous tau is a beneficial treatment to protect neurons from Aβ toxicity [
76]. However, our results imply that such tau-reducing therapy would result in failure to prevent Aβ-induced spine alteration.
Although AβOs generated by the Osaka mutation are mostly intracellular, picomolar amounts are released from cells (ELISA data; this study). We found that extracellular wild-type Aβ showed a trophic effect on synapses at physiological low concentrations; however, Osaka-mutant Aβ showed no trophic effect on spines. Recent evidence suggests that endogenous Aβ modulates synaptic plasticity [
40,
41] and regulates neurotransmitter release probability [
77]. These positive effects of Aβ are observed at picomolar concentrations, and higher, nanomolar concentrations lead to toxicity. In cultured hippocampal slices, we previously observed that relatively low concentrations of wild-type Aβ enhanced the levels of synaptophysin, whereas Osaka-mutant Aβ did not [
42]. In the present study, unlike wild-type Aβ, Osaka-mutant Aβ did not show trophic effects on spines at the same low concentrations. The lack of such trophic actions may in part account for the synaptic alteration associated with APP
OSK.
Acknowledgements
We thank Kenji Fukunaga, Atsushi Koyama, Reina Fujita, and Maiko Mori for technical assistance and Kathlyn Gan for critical comments on the manuscript. This study was supported by the Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan (no. 23110514, 24659434, 25290018); by the Grants-in-Aid from the Ministry of Health, Labour, and Welfare, Japan; and in part by the Strategic Research Program for Brain Sciences (CREST), the Ministry of Education, Culture, Sports, Science and Technology of Japan, and the Canadian Institute for Health Research (no. 90396) to M.A.S. For initiating this work, M.A.S. thanks the Japanese Society for Promotion of Science Long-Term Invitation Fellowship (#L11710).
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
MAS and TT designed experiments. TU, EMR, MY, KN and MAS performed experiments and analyzed data. MAS and TT wrote the manuscript. All authors read and approved the final manuscript.