Early Endosomal Abnormalities and Cholinergic Neuron Degeneration in Amyloid-β Protein Precursor Transgenic Mice

Abstract

Early endosomal changes, a prominent pathology in neurons early in Alzheimer’s disease, also occur in neurons and peripheral tissues in Down syndrome. While in Down syndrome models increased amyloid-β protein precursor (AβPP) expression is known to be a necessary contributor on the trisomic background to this early endosomal pathology, increased AβPP alone has yet to be shown to be sufficient to drive early endosomal alterations in neurons. Comparing two AβPP transgenic mouse models, one that contains the AβPP Swedish K670N/M671L double mutation at the β-cleavage site (APP23) and one that has the AβPP London V717I mutation near the γ-cleavage site (APPLd2), we show significantly altered early endosome morphology in fronto-parietal neurons as well as enlargement of early endosomes in basal forebrain cholinergic neurons of the medial septal nucleus in the APP23 model, which has the higher levels of AβPP β-C-terminal fragment (βCTF) accumulation. Early endosomal changes correlate with a marked loss of the cholinergic population, which is consistent with the known dependence of the large projection cholinergic cells on endosome-mediated retrograde neurotrophic transport. Our findings support the idea that increased expression of AβPP and AβPP metabolites in neurons is sufficient to drive early endosomal abnormalities in vivo, and that disruption of the endocytic system is likely to contribute to basal forebrain cholinergic vulnerability.

INTRODUCTION

Neuronal early endosome alterations occur early in Alzheimer’s disease (AD), preceding amyloid-β (Aβ) accumulation in the frontal cortex and robust neurofibrillary tangle pathology [1]. While the etiology of AD-related neuronal early endosomal dysfunction is likely to be complex, dependent in part upon lipid metabolism [2-4], apolipoprotein E allelic variation [1] as well as other genetic risk factors [5], growing evidence suggests that amyloid-β protein precursor (AβPP) levels and the levels of specific AβPP metabolites may contribute to endosomal disruption in neurons [6] and non-neuronal cells [7]. In human Down syndrome (DS), in which the gene for App is triplicated, neuronal early endosomal changes similar to those seen in AD occur decades prior to Aβ pathology [1], and are additionally seen in non-neuronal cells such as cultured DS fibroblasts [8]. In the DS mouse model Ts65Dn, neuronal early endocytic morphological disruption is dependent upon App gene triplication, arguing that increased AβPP expression is necessary for early endosomal disruption in the context of DS trisomy [6]. Recently, we and our colleagues have shown that AβPP overexpression, and in particular the overexpression of the β-cleavage-generated C-terminal fragment (βCTF) of AβPP, is necessary for the manifestation of early endosomal changes in cultured DS fibroblasts [7]. Consistent with the idea that AβPP overexpression itself can lead to early endosomal disruption, increased expression in the cortex of endocytic proteins, such as clathrin and dynamin, has been reported in the AβPP-overexpressing Tg2576 mouse, although early endosome morphology was not directly examined [9]. In our study, we set out to examine endosomes in two mouse models with AβPP overexpression, the APP23 mouse model that has the AβPP Swedish K670N/M671L double mutation at the β-cleavage site resulting in increased production of βCTFs, and the APPLd2 mouse model that has the AβPP London V717I mutation near the γ-cleavage site, resulting in increased Aβ₄₂ production [10-12].

Basal forebrain cholinergic neurons (BFCNs) are a widely examined neuronal subpopulation that are dependent upon early endosome-mediated signaling [13, 14] and are particularly vulnerable to degeneration in AD [15-19]. These neurons, with long axonal projections to distal sites, are dependent on proper endosome function to deliver critical nerve growth factor retrograde trophic-support via signaling endosomes from the axon terminus to the cell soma [14, 20-22]. In addition to deficient neurotrophic factor signaling leading to neurodegeneration, altered early endocytic function in neurons is likely to have multiple pathological consequences for the cell [23, 24], including disruption of AβPP metabolism [25].

Mouse models of AβPP overexpression and Aβ accumulation typically show some levels of BFCN changes. There are prior reports of cholinergic fiber loss and distortion in APP23 mice [26] and in APPLd2 mice [27]. Previous studies have also reported smaller BFCN cell bodies in APP23 medial septal nucleus (MSN) [26], consistent with cholinergic system impairment in APP23 mice as demonstrated by reduced acetylcholinesterase and choline acetyltransferase (ChAT) enzyme activity [26, 28]. These changes in BFCN morphology and cholinergic biochemistry have not yet been shown to be associated with the endocytic system changes, nor has endosome morphology been characterized in these AβPP overexpression mouse models. Thus, we set out to characterize these systems in the APP23 and APPLd2 mice.

MATERIALS AND METHODS

Animals

All mouse experimentation and animal care was conducted in accordance to protocols approved by the Nathan S. Kline Institute’s Institutional Animal Care and Use Committee. APPLd2 mice (n = 15) and APP23 mice (n = 14) were maintained on a C57BL/6J background, and wild-type (wt) littermates of both lines (n = 16) were used as control animals. Animals of both sexes ranging in age 12–14 months were used in this study.

Immunolabeling and antibodies

For all immunohistological procedures, mice were anesthetized with ketamine (50 mg/kg)/xylazine (5 mg/kg) and transcardially perfusion-fixed with 4% paraformaldehyde in 0.1 M sodium cacodylate buffer. Brains were removed and post-fixed overnight in 4% paraformaldehyde in phosphate buffered saline (PBS) at 4°C, transferred to PBS the next day, and subsequently cut into 40 μm-thick coronal sections with a vibratome. Free-floating sections from all the mouse groups were concurrently processed for immunohistochemical examination [29]. Control sections were processed with the omission of the primary or secondary antibodies to exclude non-specific reactions.

Using goat polyclonal antibodies, fluorescent labeling was performed to label rab5b (sc-26569; Santa Cruz Biotechnology, Santa Cruz, CA; dilution 1 : 100) or rabaptin-5 (sc-6162; Santa Cruz Biotechnology, Santa Cruz, CA; dilution 1 : 100). Double immunofluorescence labeling was performed to identify coincidence of AβPP and ChAT using either monoclonal anti-AβPP/βCTF/Aβ antibody 6E10 (SIG-39320; Convance ImmunoTechnologies, Denver, PA; dilution 1 : 200) or monoclonal antibody C1/6.1 (1 μg/ml, which recognizes the carboxyl-terminal cytoplasmic domain of AβPP [25]) with goat polyclonal anti-ChAT antibody (AB144P; EMD Millipore, Billerica, MA; dilution 1 : 500). ChAT colocalization with rab5b was determined by dual immunostaining using rabbit anti-ChAT (AB143; EMD Millipore, Billerica, MA; dilution 1 : 500) and the earlier described goat anti-rab5b antibody. Following binding of fluorophore-conjugated secondary antibody (Invitrogen, Grand Island, NY; dilution 1 : 500), immunofluorescence was observed and captured at 100× magnification using an LSM 510 Meta confocal microscope (Zeiss, Thornwood, NY). In total, approximately 30 neurons per animal were measured (n = 4 per genotype) randomly by a genotype-blinded observer. Appropriate anatomical regions were identified at low-power magnification by a blinded observer. Switching to the 100× objective, the center-field neuron(s) were analyzed. Each neuron sampled was at least two cell-lengths from the closest sampled neuron, and neurons from several sections throughout the region were analyzed, totaling 10–12 neurons from 3 brain sections per animal. Quantitation of rab5b signal was measured using ImageJ (NIH, USA) after thresholding the density of rab5b-positive signal over background; the average endosomal size was calculated as the ratio of positive-pixels per rab5 positive endosomes in a neuron to the number of endosomes in the neuron.

For stereological counting of ChAT neurons, diaminobenzidine (DAB) staining was employed. Following incubation in anti-ChAT (AB144P) primary antibody, sections were washed in diluting buffer and treated with biotinylated rabbit anti-sheep IgG antibody (Vector Laboratories, Burlingame, CA; 1 : 500), and developed using the avidin-biotin-peroxidase solution (ABC kit, Vector Laboratories, Burlingame, CA) and DAB (Sigma, St. Louis, MO).

Optical fractionator procedure

The optical fractionator system consisted of a computer assisted image analysis system including a Zeiss Axioskop II microscope hard-coupled to an Applied Scientific Instrumentation MS-2000 computer controlled x–y–z motorized stage, a Qimaging Qicam Fast 1394 color camera, a Dell Optiflex 755 computer, and Bioquant Life Science stereological software (Bioquant Image Analysis, Nashville, TN). Stereological cell counting was conducted as previously described by Granholm et al. [30]. The MSN was outlined under a low magnification (2.5×) using the following land-marks: The rostral border consisted of the medial orbital cortex (at the level of the midline fusion of the corpus callosum), the caudal border consisted of the midline fusion of the anterior commissure, and the lateral borders consisted of the shell of the Accumbens nucleus. The outlined region was overlaid by a systematic random design of dissector counting frames (100 × 100 μm), and a 40× objective lens with a 0.75 numerical aperture was used to count and measure individual cells within the counting frames. Every third section containing the MSN was evaluated and quantitative estimates of the total number and cell-soma size of ChAT-immunoreactive neurons in the MSN was calculated using this unbiased, stereological cell counting method [30, 31]. Cell volume was calculated using rotator probe stereological methods provided by the Bioquant software. Minimally, 100 cells were counted from each brain in a double-blind manner. The counting brick was 20-μm thick. The analysis rendered a mean coefficient of error of 0.05–0.10 as calculated according to Gundersen et al. [32]. An upper guard zone of 2 μm was excluded from counting.

Western blot analysis

For western blot analyses, mice were euthanized and brains were immediately dissected and frozen on dry ice. Frozen hemibrains were homogenized as previously described using protease inhibitors [33], and proteins separated by 4–20% Tris-HCl SDS-PAGE and transferred to polyvinylidene difluoride membranes. The anti-AβPP C-terminal monoclonal antibody C1/6.1 (1 μg/ml) recognizes both mouse and human APP holoprotein and C-terminal fragments (CTFs) [25, 33, 34]. Membranes were incubated overnight in primary antibody, washed, and incubated with horseradish peroxidase-conjugated goat anti-mouse IgG (MP Biomedicals, Irvine, CA; dilution 1 : 5000) for 1 h. ECL substrate (Amersham Biosciences, Piscataway, NJ) was added before exposure to x-ray film. The membrane was then stripped and reprobed with anti-β-tubulin antibody (Abcam, Cambridge, MA; dilution 1 : 5000) as a loading control. Blots were quantified using ImageJ (NIH, USA) and the density of signal was normalized against β-tubulin.

Statistical analysis

One-way analysis of variance (ANOVA), followed by posthoc multiple comparison using Dunnett’s test, was used to evaluate the differences between genotypes. Error bars represent standard error of the mean.

RESULTS

AβPP overexpressing transgenic mouse models have altered early endosome morphology

We labeled mouse brain tissue with an antibody against the small GTPase rab5b, a regulator of endocytosis and early endosome fusion and a specific marker for early endosomes [6]. Rab5b-positive earlyendosomes in frontoparietal cortical neurons from APP23 mice (middle panel, top row, Fig. 1A) and APPLd2 mice (right panel, top row, Fig. 1A) appeared abnormally large compared to the neurons from wt mice (left panel, top row, Fig. 1A), consistent with previous findings that early endosomal abnormalities in non-neuronal cells can be AβPP-dependent [7]. Rabaptin-5, which is an effector-protein of rab5, has also been used by our group to characterize early endosomes in brain tissue [6]. We found a robust increase in the expression of rabaptin-5 in both APP23 (middle panel, bottom row, Fig. 1A) and APPLd2 mice (right panel, bottom row, Fig. 1A) when compared to wt mice (left panel, bottom row, Fig. 1A).

Fig. 1.

The morphology of early endosomes is altered in AβPP overexpressing transgenic mice. (A) Neurons from the frontoparietal cortex of APP23 (middle panel) and APPLd2 (right panel) mice labeled with rab5b (top row) and rabaptin-5 (bottom row) show enlarged endosomes and increased immunoreactivity compared to age-matched wt controls (left panel). As described in the Materials and Methods section, quantitation of average rab5b-positive early endosome size in APP23 and APPLd2 compared with wt mice is shown in (B). Measurements were made from 10–12 neurons per section, 3 sections per mouse, and 4 mice per genotype. Labeling conditions and exposure times were identical throughout. (C) Western blot of proteins from hemibrains lacking cerebella of wt, APP23, and APPLd2 mice probed with an anti-AβPP C-terminal monoclonal antibody (C1/6.1), which detects both murine and human AβPP and CTFs (upper and middle panels). β-tubulin reactivity is shown as a loading control (lower panel). Quantification of AβPP (D) and CTF (E) band density (arbitrary units normalized to the β-tubulin bands, mean ± SEM). Scale bars, top row 10 μm, bottom row 100 μm. Differences from wt were significant at *p < 0.05, **p < 0.01, and ***p < 0.001.

Rab5b-immunolabelling of early endosomes in randomly chosen cortical neurons was quantitated, with the average size of individual endosomes compared among the three genotypes. Early endosomes in APP23 mice were found to be 2.42 ± 0.50 (p < 0.01) times the average size of rab5b-positive endosomes in wt mice (Fig. 1B), while APPLd2 mice showed a non-significant trend toward increased average early endosomal size (1.52 ± 0.31 times wt) (Fig. 1B).

AβPP expression and early endosomal morphology in BFCNs in the MSN of APP23 and APPLd2 mice

Robust global overexpression of AβPP and AβPP CTFs in both APP23 and APPLd2 mice was seen by western blot from hemibrain homogenates of 12-month-old mice, with holoprotein levels greater in the APPLd2 mouse (Fig. 1C, D). As predicted given the β-cleavage promoting Swedish mutation expressed in the APP23 mouse [35], the increase in βCTF levels was found to be greatest in this line (Fig. 1C, E). The more marked endosomal changes in the APP23 model are in agreement with the idea that βCTFs are the primary mediators of AβPP-driven endosome disruption [7]. In order to examine AβPP- and AβPP metabolite-driven endosomal changes in cholinergic neurons of the MSN, we first confirmed that human AβPP is overexpressed in cholinergic neurons of the transgenic mice by double immunolabeling with antibodies against human AβPP/βCTF/Aβ (6E10; Fig. 2A) or human and murine AβPP/CTFs (αCTF + βCTF) but not Aβ (C1/6.1; Fig. 2B). Both APP23 (middle panels, Fig. 2A, B) and APPLd2 (bottom panels, Fig. 2A, B) showed robust human AβPP expression in cells of the MSN, including ChAT-positive cholinergic neurons. That both antibodies showed a similar cellular immunolabeling pattern confirms that BFCNs overexpress AβPP in both transgenic models and that the immuno-signal is not only due to Aβ immunoreactivity.

Fig. 2.

Basal forebrain cholinergic neurons from APP23 and APPLd2 mice overexpress human AβPP. A) Co-immunolabeling for ChAT (red) and antibody 6E10 (green), which reacts with human AβPP, βCTFs, and Aβ, in the MSN of APP23 (middle panels) and APPLd2 (bottom panels) mouse brain tissue. No 6E10 immunolabeling is seen in the wt mice (top panels). B) Co-immunolabeling for ChAT (red) and antibody C1/6.1 (green) [25], which reacts with human and murine AβPP/αCTF/βCTF, in the MSN of APP23 (middle panels) and APPLd2 (bottom panels). Endogenous murine AβPP labeling is seen in the wt mouse tissue (top panels). Scale bars 100 μm.

Changes in endosomal morphology in these BFCNs were demonstrated by double immunolabeling with antibodies against rab5b and ChAT (Fig. 3A). As we did for cortical neurons, we compared the size of rab5b-positive endosomes in ChAT-positive MSN cholinergic cells. Cholinergic early endosomal size in APP23 mice was significantly increased when compared to wt (1.42 ± 0.12% increase over wt, p < 0.05), while APPLd2 again showed a non-significant trend toward an increase in early endosomal size (1.2 ± 0.08% increase over wt) (Fig. 3B).

Fig. 3.

Early endosomal disruption in basal forebrain cholinergic neurons. A) BFCNs in the MSN of APP23 and APPLd2 mice were identified using an antibody against ChAT (green), with rab5b immunolabeling used to show early endosomal morphology (red). Quantitation of average endosome size, as described in the Materials and Methods section, is depicted in (B) (average endosome size ± SEM). Individual endosome area from a minimum of 100 cholinergic neurons per genotype was measured (n = 4 each genotype). Labeling conditions and exposure times were identical throughout. Scale bar, 10 μm. Differences from wt were significant at **p < 0.01.

Cholinergic neuron loss in the MSN of APP23 and APPLd2 mice

We next examined the morphology of cholinergic neurons and fibers in the MSN. Compared to wt ChAT-positive cells, ChAT-positive cells in both APP23 and APPLd2 displayed dystrophic neurites, with fiber deafferentation and retracted neurites (Fig. 4A). Additionally, ChAT-positive cells in AβPP transgenic mice were pyknotic compared to wt mice ChAT-positive cells (Fig. 4A). Using quantitative unbiased stereology, we found that both APP23 and APPLd2 mice (n=7 and n = 8, respectively) had significantly fewer ChAT-positive cells in the MSN (3,119 ± 122; 2,716 ± 422, respectively) compared to wt mice (n = 6; 4,293 ± 402; p < 0.05 for both transgenic lines) (Fig. 4B). BFCNs in APP23 mice also exhibited smaller average cell soma volume (1,107 ± 26 μm³) when compared to wt mice (1,542 ± 136 μm³; p < 0.01), whereas APPLd2 mice (n = 8) did not show a similar atrophy in BFCN soma size (1,598 ± 96 μm³) (Fig. 4C).

Fig. 4.

Cholinergic neuron degeneration in the MSN of APP23 and APPLd2 mice. A) ChAT-immuno-positive neurons in the MSN of APP23 (middle panel, n = 7) and APPLd2 (right panel, n = 8) show fiber deafferentation and retracted neurites compared to age-matched wt controls (left panel, n = 6). Unbiased stereological quantitation of ChAT-immuno-positive cell number (B) and cell-soma volume (C) are depicted (cell number ± SEM; mean volume ± SEM) (wt, n = 6; APP23, n = 7; APPLd2, n = 8). Scale bar, 50 μm. Differences from wt were significant at *p < 0.05; **p < 0.01.

DISCUSSION

We have examined two mouse models with neuronal AβPP overexpression to assess neuronal early endosome morphology, finding that the βCTF accumulating APP23 model, and to a lesser extent the APPLd2 mouse, develops early endosomal alterations that closely resemble those seen early in the pathogenesis of the disease in AD patients [1, 36, 37]. Similar to the human studies in which neuronal endosome changes were characterized in AD patients prior to florid plaque deposition [1, 36, 37], our studies parallel the human AD findings in that the APP23 and APPLd2 mice have limited Aβ deposition at 12 months of age [35, 38], yet show early endosomal pathology. Our group and others have used DS systems, both in vitro and in vivo, as a model of AD-related early endosomal changes in addition to studying DS pathobiology [6-8, 39]. In murine DS models, Aβ accumulation does not occur but endosomal disruption is prominent [7, 33]. While AβPP overexpression appears to be a necessary contributor to endosomal alterations in these in vivo and in vitro DS models [7, 39, 40], it has not been previously shown in neurons that early endosomal alterations can be driven solely by AβPP overexpression without the contributing effects from other DS-region triplicated genes. Indeed, our earlier study of brain AβPP expression in the Ts65Dn mouse model shows that endosomal alterations occur prior to an increase in AβPP mRNA and protein levels in the whole brain [6, 33], suggesting that DS trisomic cells may be particularly vulnerable to early endosomal disruption. Our findings disassociate the role of DS-related triplication of genes to endosomal abnormalities previously described; arguing instead that sufficient overexpression of AβPP can drive neuronal early endosomal changes in vivo, thus extending this pathological alteration to commonly used AD mouse models.

Recent cell culture studies have directly implicated the cell-associated βCTF of AβPP in the development of endosomal pathology in cultured non-neuronal cells, including a β-cleavage-dependence in DS cells for this endosomal phenotype [7]. The greater endosomal pathology seen in the APP23 mouse as comparedto APPLd2 mice correlates with the robust expression of βCTFs in this model. Although the possibility cannot be ruled out that early Aβ deposition also affects neuronal endocytosis, previous evidence has shown that endosome pathology is not Aβ-dependent both in vitro [7] and in vivo in BFCNs of a DS mouse model [39], strongly suggesting that AβPP and cell-associated AβPP metabolite levels contribute to early endosomal disruption.

The loss of BFCN and deficiencies in ChAT cholinergic enzyme activities in the pathogenesis of both DS and AD make this subpopulation of neurons of particular interest [15, 41-44]. Of varying severity and magnitude depending, in part, on age and brain region, many AβPP overexpressing mouse models have been shown to exhibit cholinergic deficits, including APP23 [26, 28], APPLd2 [27], TgCRND8 [45], Tg2576 [46], PDAPP [47], PDAPP_(Sw,Ind) [48], and APP/PS1KI [49] mice. In addition to dystrophic neurites, we demonstrate a significant loss of cholinergic cells in the MSN of APP23 and APPLd2 mice that parallels the findings in the brains of DS and AD patients [15, 41]. Others have attributed this cholinergic fiber loss with Aβ load [50, 51], and it had been postulated that cortical cholinergic deficit in APP23 mice is a locally induced phenomenon driven by Aβ deposition in the MSN [26]. Our findings suggest an alternative and additional explanation consistent with the DS mouse model, that this vulnerable neuronal population has altered early endosomes in AβPP overexpressing mice. These neurons have long projections into the hippocampus and cortex of the brain, and are depe dent on retrograde trophic signaling to maintain cell viability and phenotype [14, 20, 22]. In this population, signaling endosomes mediate axon-to-cell-soma transport of nerve growth factor (NGF), and studies in a mouse model of DS have shown that abnormalities in these “signaling” endosomes that compromise NGF transport are App-dosage dependent [6, 39]. Our findings provide further support for the idea that altered endocytosis is a mechanism by which BFCNs degenerate in AβPP overexpression mouse models. Consistent with a key role for the βCTF in driving early endosomal changes [28], the APP23 mouse with the βCTF-generation promoting Swedish mutation has greater neuronal early endosome alterations and BFCN pathology, with significantly reduced cholinergic cell volume compared to the APPLd2 mouse.

The vast majority of AD cases are sporadic and late-onset [52], and while increased expression of AβPPhas been suggested to be a contributing factor to the development of sporadic AD [53, 54], extensive overexpression of AβPP, as occurs in the AβPP transgenic mice and with aging in the DS mouse models [13], is unlikely in the sporadic human disease. Nevertheless, endocytosis-dependent synaptic activity has been shown to increase Aβ production, apparently by promoting βCTF generation [55, 56]. Other risk factors for late-onset AD, such as altered lipid levels, may similarly increase the generation of βCTFs within neurons [2, 3, 57], suggesting that multiple mechanisms relevant to AD pathogenesis and Aβ production may concurrently increase cellular βCTF levels. While the increase in neuronal βCTF levels due to suchsporadic-risk factors may be moderate, prior findings in the DS model argues that limited changes in AβPP levels are sufficient to lead to robust and pathologically significant early endosomal changes [7], presumably because of the vulnerability of the cells to these changes in the trisomic background. The overlaying of multiple factors in AD that are capable of altering endocytic function, including the direct impact of lipids and cholesterol, genetic vulnerabilities, and perhaps aging, can all contribute to an endocytic system that is poised to disruption by moderate changes in βCTF levels. Abnormal endocytic function is likely to have many pathological consequences for neurons [23, 24], and the growing consistency that such changes are detected in multiple cell types and multiple models in which AβPP and βCTF expression is increased highlights the importance of AβPP and its processing to this system.