Background
A transmembrane aspartyl protease, termed β-site APP cleaving enzyme 1 (BACE1), initiates Alzheimer's disease (AD) β-amyloid (Aβ) production by sequential proteolysis of amyloid precursor protein (APP) [
1‐
5]. The release of Aβ from its precursor involves initial processing by BACE1 to release the APP ectodomain, followed by intramembrane proteolysis by γ-secretase [
6]. Aβ is the major component of cerebral amyloid plaques in brains of aged individuals and those with AD. Several lines of evidence suggest that Aβ plays a central role in AD pathogenesis. For example, familial AD-linked mutations near the amino terminus of the Aβ region in APP, found in two Swedish families, cause AD by significantly increasing Aβ production due to enhanced BACE1 cleavage of APP [
7‐
9]. Moreover, a single amino acid substitution adjacent to the BACE1 cleavage site of APP, which significantly reduces BACE1 cleavage and Aβ peptide generation in cultured cells, has been recently found to protect against disease onset as well as cognitive decline in the elderly without AD [
10].
APP and BACE1 are type I transmembrane proteins that undergo secretory and endocytic trafficking. However, in cultured cell lines and primary neurons, only a subset of full-length APP is processed to generate Aβ. This implies either BACE1 cleavage of APP is rather inefficient or that BACE1 has limited access to APP due to their distinct intracellular itineraries and/or spatially restricted localization in intracellular organelles. Over the years, non-neuronal cells were used as experimental systems to characterize the cellular organelles and sorting pathways involved in amyloidogenic processing of APP. Most studies recognize the importance of endocytic trafficking of APP for Aβ production [reviewed in [
6,
11,
12]. This notion is in agreement with the predominant steady-state localization of BACE1 in endocytic organelles, and its trafficking and recycling between the plasma membrane and endosomes [
1,
13,
14]. Moreover, BACE1 activity has a low pH optimum
in vitro[
1], supporting a model where APP cleavage by BACE1 occurs in acidic intracellular organelles such as the endosomes. Indeed, siRNA knockdown of Golgi-associated, γ-adaptin homologous, ADP-ribosylation factor-interacting proteins enhance BACE1 localization in early endosomes and a concomitant increase in Aβ secretion [
15‐
17]. Finally, membrane targeting of a BACE1 transition-state inhibitor to endosomal organelles by linking it to a sterol moiety efficiently reduced BACE1 activity as compared with the free inhibitor [
18].
In the mouse brain, ~70% of Aβ released into the interstitial fluid requires ongoing endocytosis, and synaptic activity regulates the vast majority of this endocytosis-dependent Aβ production [
19,
20]. The underlying mechanism likely involves activity-dependent regulation of endocytic trafficking of APP as well as its secretases [
21]. Neuronal endosomes are not only found in the soma but are distributed throughout the dendrites and axons where they undergo bidirectional transport, adding to the complexity of the trafficking mechanisms [
22]. Indeed, APP undergoes BACE1-mediated cleavage during anterograde axonal transport, and Aβ can be generated and released at or near presynaptic sites
in vivo[
19,
23‐
27]. BACE1 has been reported to localize in dendrites and axons in cultured neurons and in the brain [
21,
28‐
32]. Axonal BACE1 localization is significant because abnormal accumulation of BACE1 in axon terminals has been documented in the brains of individuals afflicted with AD. This later finding raises the possibility that local elevation in BACE1 processing could contribute to amyloid burden in AD [
30,
33]. However, the molecular mechanisms responsible for axonal sorting of BACE1 have not been fully explored.
Here, we used live-cell imaging to characterize dynamic BACE1 transport in hippocampal neurons in vitro and in brain slices in situ. We report BACE1 colocalization and dynamic transport in recycling endosomes within the dendrites and axons of cultured hippocampal neurons. Interestingly, our results show that efficient axonal sorting of BACE1 requires Rab11 activity. Together, these findings provide the first demonstration of dynamic BACE1 transport in hippocampal neurons in situ and the involvement of neuronal recycling endosomes in axonal sorting of BACE1.
Discussion
In this study, we generated transgenic mice expressing BACE1-YFP, and for the first time, visualized BACE1 dynamic axonal transport in situ within the mossy fibers of the hippocampus by multiphoton microscopy. Moreover, we investigated the localization and trafficking of BACE1 by live-cell imaging in cultured hippocampal neurons. The dynamic characteristics of BACE1 transport in hippocampal slices and in mature primary cultured hippocampal neurons (DIV12-14) shared several similarities. In both cases we observed two distinct pools of BACE1: a highly dynamic pool of BACE1 found in tubulo-vesicular carriers that were transported bi-directionally and consistent with a microtubule-based transport mechanism; moreover, a second pool of BACE1 found in stationary structures that were relatively larger in size as compared with the motile carriers. FM-dye uptake studies revealed that stationary BACE1-YFP fluorescence corresponds to active presynaptic terminals. Finally, using immunofluorescence labeling and dual-color imaging, we demonstrate BACE1-YFP localization and dynamic transport in Rab11-positive recycling endosomes. Expression of a dominant-negative Rab11 mutant causes accumulation of internalized BACE1 in the soma concomitant with a loss of BACE1 levels in the axons, consistent with BACE1 transcytosis from the somato-dendritic compartments to the axons in endosomes.
In human brain, BACE1 can be observed by immunostaining in dendrites of CA1 neurons [
34]. Interestingly, in mouse brain, endogenous BACE1 is highly enriched in hippocampal mossy fiber terminals, and only low levels of BACE1 can be detected in the neuronal soma and dendrites (Figure
1) [
33,
34]. These apparent differences likely represent the balance between the biosynthetic level and the efficiency of the transport machinery responsible for BACE1 trafficking. Thus, overexpression of BACE1-YFP in cultured hippocampal neurons allows us to appreciate dynamic sorting of BACE1 between dendrites and axons, which is not possible to discern from steady-state analysis of endogenous proteins that have reached their preferred final destination. Indeed, our results on BACE1 localization are analogous to what has been documented for the localization of kainate-type glutamate receptor subunits GluK2 and GluK3. When analyzed by immunohistochemistry, the distribution of GluK2 and GluK3 is prominent at hippocampal mossy fiber synapses in mouse brain [
44]. However, detailed analysis of these receptors by transfection in cultured hippocampal neurons reveals a prominent dendritic localization and demonstrates that the steady-state localization of kainate receptors is achieved by dynamic activity-dependent polarized endocytic sorting from the dendrites through recycling or degradative pathways [
45,
46]. The progressive increase in BACE1 levels in axons concomitant with its decrease in dendrites over time following transfection suggest the existence of saturable mechanisms that mediate polarized axonal sorting of BACE1 (Figure
3). Axonal sorting and presynaptic localization of BACE1 are highly relevant because of the recent identification of several presynaptic proteins such as L1, CHL1, LRRN1, brain EGF-repeat containing transmembrane protein, and neurexin-1a as neuronal BACE1 substrates [
47,
48]. Moreover, loss of BACE1 expression results in axon guidance defects in the hippocampus and the olfactory bulb [
36,
44,
48].
Although the machinery and the molecular mechanisms governing polarized sorting of proteins in neurons needs to be further explored, the involvement of three pathways for the axonal targeting of transmembrane proteins has been suggested: 1) direct polarized delivery of nascent proteins from the secretory pathway to the dendrites or axons in TGN-derived vesicles, 2) non-polarized delivery to dendrites and axons followed by selective retrieval and retention involving endocytosis and degradation pathways, 3) indirect polarized delivery via endosomes, also called transcytosis, whereby nascent proteins are first delivered to the somato-dendritic cell surface, and then following endocytosis get routed to the axons [
22]. The extensive localization of BACE1 in endosomes and the highly polarized axonal localization of BACE1 in hippocampal mossy fibers ([
34], Figures
1 and
5) suggest that the direct polarized delivery is unlikely to play a major role in BACE1 axonal targeting. The results detailed in this study suggest that BACE1 sorting to Rab11-positive recycling endosomes plays an important role for further transport to the axons and presynaptic terminals. First, BACE1 is efficiently sorted in endosomes positive for three known markers of recycling endosomes: Rab11, syntaxin 13, and TfR [
21,
34]. Second, the dynamic characteristics of BACE1 transport in hippocampal neurons, characterized by live-cell imaging, are consistent with protein trafficking in recycling endosomes [
37], a conclusion supported by two-color imaging of BACE1 co-transport with Rab11 in dendrites and axons (Figure
5, Additional file
4). Third, the impairment of Rab11 activity by dominant-negative mutant expression caused internalized BACE1 accumulation in the soma with a concomitant decrease in axons. The accumulation of internalized BACE1 in the soma and unperturbed relative subcellular distribution in endosomes and lysosomes (Figure
7) suggest that BACE1-containing vesicles are not rerouted for lysosomal degradation when a block in recycling itinerary is imposed by DN Rab11 expression. Because we have not found convincing evidence for the involvement of degradative pathways to explain the paucity of BACE1 in soma and dendrites, we hypothesize that proper BACE1 sorting to axons and presynaptic terminals requires its transcytosis in a Rab11-dependent manner rather than aspecific sorting followed by selective retention/retrieval mechanisms. Interestingly, BACE1 shows only partial colocalization with Rab11 in axons, when observed by immunofluorescence staining or dual-color live imaging (Figure
5). These findings suggest that in axons a subset of BACE1 is sorted independent of Rab11-positive endosomes or Rab11 detaches from a subset of BACE1-positive vesicles. It will be of interest to identify other trafficking adaptors that regulate dynamic BACE1 transport in axons. In this regard, we recently reported that loss of Eps15-homology-domain containing 1 and 3 protein function in hippocampal neurons compromises dynamic axonal transport and overall BACE1 levels in axons [
34].
Thus far, only a few proteins have been characterized to undergo neuronal somaodendritic-to-axonal transcytosis; notable examples are the axonal cell adhesion molecule L1/NgCAM [
49], tropomyosin-related kinase Trk receptors [
41], and Contactin-associated protein 2, Caspr2 [
50]. In each case, it is clear that endosomal sorting is required for axonal targeting, but the machinery and the molecular players involved in achieving polarized sorting are not fully understood. Also, differences could be already found between the steady-state localization of these proteins suggesting the existence of different regulatory mechanisms in the transcytotic pathway. For example, L1/NgCAM axonal surface expression is prominent and highly polarized in transfected cultured neurons [
49], which is not the case for BACE1. In addition, L1/NgCAM is sorted mainly in NEEP21-positive vesicles in soma/dendrites before transcytosis, and shows only low levels of colocalization with transferrin-positive recycling endosomes. In contrast, a large fraction of BACE1 is found in recycling endosomes positive for TfR and Rab11, with only a relatively smaller fraction present in EEA1 or NEEP21-positive early endosomes (Figure
5, [
34]). Thus, it appears that L1/NgCAM and BACE1 are subject to different the endosomal sorting steps prior to transcytosis. BACE1 sorting in recycling endosomes and dynamics share similarities with Trk receptors. Similar to BACE1 trafficking in hippocampal neurons, TrkA is also dynamically co-transported in axons in Rab11-positive vesicles and its trafficking in sympathetic neurons also requires Rab11-GTPase activity for efficient transcytosis [
41]. It has been hypothesized that transcytosis serves as a means to facilitate efficient axonal delivery of receptors in response to signals often involving ligand:receptor interaction in presynaptic terminals as it is shown for TrkA transcytosis mediated by nerve growth factor signaling [
41]. Since BACE1 is the first reported enzyme to undergo transcytosis, exploring potential signals that regulate BACE1 transcytosis would be of great interest especially for a better understanding of the mechanism underlying its abnormal accumulation in swollen presynaptic terminals in Alzheimer’s disease. Since synaptic activity regulates BACE1 trafficking and APP processing [
19‐
21] it is possible that synaptic activity impairment observed in Alzheimer’s disease could affect BACE1 transcytosis in a manner that promotes its accumulation in dystrophic neurites, thus contributing to local Aβ production near the presynaptic terminals.
Abnormalities of the endosomal system such as the fusion of early and recycling endosomes have long been implicated in Alzheimer’s disease pathogenesis (reviewed in [
11]). Interestingly, Rab11 was recently identified as one of the two major regulators of Aβ production, in an unbiased RNAi screen of human Rab GTPases and Rab GTPase-activating proteins by Rajendran and colleagues [
51]. Overexpression of dominant-negative mutants of Rab11a or Rab11b also reduced Aβ and sAPPβ levels significantly. Moreover, siRNA knockdown of Rab11 expression in primary neurons significantly reduce sAPPβ and Aβ levels demonstrating that Rab11 function is crucial for β-cleavage and Aβ generation [
51]. Our characterization of Rab11 as a novel regulator of BACE1 axonal sorting in neurons, along with the identification of Rab11 as a modulator of Aβ production, raises the possibility that dysfunction of Rab11 may underlie pathogenesis in a subset of sporadic Alzheimer’s disease cases. Indeed, aberrant Rab11 trafficking has been reported in Huntington's disease and contributes to oxidative stress and neuronal cell death [
52]. Future studies aimed at manipulating of polarized neuronal BACE1 trafficking to assess potential modulation of Aβ production
in vivo would be necessary to evaluate if trafficking modulation of BACE1 could serve as a therapeutic strategy in Alzheimer’s disease to reduce cerebral amyloid burden.
Methods
cDNA constructs
The expression plasmids that encode BACE1-YFP and BBS-BACE1-YFP (harboring the 13-amino acid α-Bungaratoxin Binding Site) were generated as described in [
34]. A similar strategy was used to generate BACE1-Cerulean construct. To generate mouse Rab11b expression plasmids, a mouse brain PCR product that codes for amino acids 72 to 218 was exchanged for the corresponding region in HA-tagged Rab11a WT and Rab11a
S25N constructs [
53]. The cDNA inserts were then subcloned in-frame into the pmCherry-C1 vector. All constructs were verified by sequencing. pCS2-Cerulean was generated by subcloning the Cerulean coding sequence downstream of the CMV promoter. The Cerulean-tagged synaptophysin expression construct was provided by Dr. Karen L. O'Malley.
Generation of BACE1-YFP transgenic mice
BACE1 transgenic mice were generated in which expression of BACE1-YFP is regulated by tetracycline [Tet-off system] [
54]. A cDNA that encodes human BACE1 fused with YFP through a 22-residue linker was subcloned into the
EcoRV site of the
tetO promoter expression vector pMM400 (gift of M. Mayford, The Scripps Research Institute, La Jolla, CA). The
tetO promoter BACE1-YFP transgene was excised with
NotI, purified and microinjected into single mouse embryos (Transgenic and Targeted Mutagenesis Lab, NU).
tetO promoter-BACE1-YFP transgenic lines were established and bred with CAMKIIα promoter-tTA transgenic mice (line B, gift of M. Mayford). Bigenic
tetO promoter-BACE1-YFP/CamKIIα promoter tTA offspring were identified by PCR, weaned and fed either normal mouse chow or chow containing 200 mg/kg of doxycyline for 4 weeks. Expression of the transgene varied from ~1.5 to >10-fold over endogenous BACE1 (as assessed by Western blot analysis) in bigenic mice derived from different
tetO promoter-BACE1-YFP lines, but the distribution and localization of BACE1-YFP was similar. Animals from a high-expressor line (#429) was used in this study due to it producing images with the greatest signal to noise ratio.
Immunohistochemistry and immunofluorescence staining
Brains were harvested from bigenic BACE1-YFP mice (3-month) and WT littermates were fixed overnight at 4°C in 4% paraformaldehyde and cryoprotected in PBS 30% sucrose/PBS. Immunofluorescence staining was performed on 40 μm coronal sections or 30 μm sagittal sections as described [
30,
55,
56]. The following primary antibodies were used: BACE1 rabbit mAb D10E5 (1:125; Cell Signaling), synaptophysin pAb (1:75; R&D System), polyclonal MAP2 (1:250, SantaCruz), MAP2 mAb (1:500; gift of Dr. Lester Binder), and neurofilament mAb NFT160 (1:3,700; Sigma-Aldrich). Alexa Fluor 488-, 568-, or 647-conjugated secondary antibodies (Molecular Probes) were used for detection. Images were acquired on Zeiss LSM 510 META laser scanning confocal microscope (BACE1-YFP transgenic brain in Figures
1C and D) or Leica SP5 II STED-CW Superresolution laser scanning confocal microscope (Figure
1C), and processed using ImageJ software.
Live imaging on hippocampal slices
Freshly harvested brains of bigenic mice were placed in cold ACSF aerated with 95% CO2/5% O2 blood gas mixture and cut into 500 μm horizontal slices. The slices were then maintained at room temperature in a bath perfused with aerated ACSF. For live-cell imaging, each slice was positioned on the microscope stage such that the dentate gyrus was on the left side of the image, perfused with ACSF and maintained at 30°C. Slices were excited at 920 nm with Mai-Tai HP DeepSee-OL laser (Spectra-Physics, laser range 690–1040 nm) using a Olympus FV1000MPE multiphoton confocal system mounted on a BX61WI frame. Images were acquired using 60X (NA 1.1) water immersion objective at the rate of 0.8 frame/sec for 4 min.
Immunofluorescence staining and live cell imaging of primary hippocampal neurons
Hippocampal neurons were cultured from E17 mouse embryos as previously described [
34,
57]. Dissociated neurons were cultured on poly-D-lysine-coated glass coverslips suspended over a monolayer of primary astrocytes prepared from P0-P2 mouse cortex. Cultures were maintained in Neurobasal supplemented with B27 serum-free and GlutaMAX-I supplement (Invitrogen). Neurons were transfected with Lipofectamine2000 (Invitrogen) on DIV 5 and fixed at various maturation stages in 4% paraformaldehyde containing 4% sucrose. For live-cell imaging, neurons were transfected on DIV11 and used between DIV12-14. The coverslips were maintained in imaging medium (119 mM NaCl, 2.5 mM KCl, 2 mM CaCl
2, 2 mM MgCl
2, 30 mM D-glucose, and 25 mM HEPES; pH 7.4) during image acquisition. Functional synapses were labeled in live DIV13 neurons transfected with BACE1-Cerulean by allowing FM
1-43 Dye uptake, essentially as described [
58].
BTX uptake experiments to label internalized BACE1 were performed essentially as described [
34]. Briefly, AF647-conjugated BTX (Invitrogen) was added to the culture medium (6.6 μg/ml final), at 37°C for 3 to 4 h. Coverslips were washed with ice-cold Hanks' balanced salt solution with 10 mM HEPES, pH 7.3, and BTX bound to cell surface BACE1 was removed by incubation in an acidic solution (0.5 M NaCl and 0.2 M acetic acid, pH 2.8) for 2 min before fixation.
Fixed neurons on coverslips were quenched with 50 mM of NH4Cl for 10 min, permeabilized for 6 min on ice with 0.2% Triton X-100, and blocked with 3% BSA in PBS. The coverslips were incubated for 1 h at room temperature with the primary antibodies diluted in PBS containing 3% BSA: MAP2 mAb (1:5,000; Sigma), EEA1 (1:200; Millipore), Syntaxin 13 (1:1000; Synaptic Systems); TfR mAb (C2F2, 1:250; Pharmingen), rat anti-LAMP1 (ID4B, 1:150; Developmental Studies Hybridoma Bank). Subsequently, the coverslips were incubated with Alexa 555- or 647-conjugated secondary antibodies (Molecular Probes) for 1 h at room temperature and mounted using Permafluor (Thermo Fisher).
Image acquisition and analysis
Wide-field epifluorescence images of fixed neurons were acquired as 200 nm
z-stacks using 20X (NA 0.75) or 60X (NA 1.49) objectives. Confocal images were acquired on a Leica SP5 II STED-CW Superresolution Laser Scanning Confocal microscope using 10X (NA 0.4) and 100X (NA 1.4; zoom 2.5) objectives. Image stacks were deconvolved using Huygens software (Scientific Volume Imaging). Extended Depth of Field plugin of ImageJ was used to generate single plane projections from processed
z-stacks [
59]. Quantitative image analysis was performed using Metamorph (Molecular Devices) and ImageJ [
60] softwares. Axonal and dendritic BACE1 fluorescence intensities were quantified on 10X single plane images or 60X
z-stack projections of neurons using an established method as described previously [
34,
61]. Briefly, the average fluorescence intensities were measured along 100–200 μm-long 1 pixel-wide line segments traced on 2–3 representative sections of dendrites and axons in each neuron using ImageJ. The mean fluorescence intensity in the soma was quantified by drawing a region around the soma. The average axon:dendrite and cell body:dendrite ratios were calculated for each neuron. Normalized axon:dendrite ratio was then calculated by dividing the raw axon:dendrite ratio of BACE1 by axon:dendrite ratio of Cerulean in the same neuron [
36]. Manders’ coefficient of colocalization of BACE1 with organelles markers or mcherry-Rab11b was calculated on thresholded confocal 100X images of dendrites and cell bodies, or 60X deconvolved
z-stack projections of axons (identified by the exclusion of MAP2 staining) using JACoP ImageJ plugin [
62].
Live-cell imaging
Live-cell images were acquired on a motorized Nikon TE 2000 microscope maintained at 37°C in a custom-designed environment chamber, at the rate of 1 frame/sec, using 60X (NA 1.49) objective and Cascade II:512 CCD camera (Photometrics). Dual-color imaging was performed using the Dual View Imaging System (Optical Insights, LLC). Image stacks were processed using ImageJ. Kymographs were generated in Metamorph, and used to determine the frequency and directionality of particle movement, and to quantify maximum velocities of particles moving at the rate of >0.1 μm/sec.
Statistical analysis
Each experiment was performed using at least three independent sets of cultures unless otherwise specified. Data are presented as mean ± SEM. Statistical significance was determined by t-tests (two groups) or ANOVA (three or more groups) using GraphPad Prism software and indicated the figures: * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001; ns - non-significant.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
VBP, CGF, and GT designed experiments, performed live cell imaging, confocal microscopy, and analyzed the data. VBP, CGF, KSV, and JR generated expression plasmids and immunostaining. SR and RV generated and characterized BACE1-YFP transgenic mice. VBP, SR, and JW performed multiphoton imaging. VBP and GT wrote the manuscript. GT conceived of the study, coordinated data analysis, and prepared the final manuscript. All authors read and approved the final manuscript.