Neuron loss in the 5XFAD mouse model of Alzheimer’s disease correlates with intraneuronal Aβ42 accumulation and Caspase-3 activation
Although the mechanism of neuron loss in Alzheimer’s disease (AD) is enigmatic, it is associated with cerebral accumulation of Aβ42. The 5XFAD mouse model of amyloid deposition expresses five familial AD (FAD) mutations that are additive in driving Aβ42 overproduction. 5XFAD mice exhibit intraneuronal Aβ42 accumulation at 1.5 months, amyloid deposition at 2 months, and memory deficits by 4 months of age.
Here, we demonstrate by unbiased stereology that statistically significant neuron loss occurs by 9 months of age in 5XFAD mice. We validated two Aβ42-selective antibodies by immunostaining 5XFAD; BACE1−/− bigenic brain sections and then used these antibodies to show that intraneuronal Aβ42 and amyloid deposition develop in the same regions where neuron loss is observed in 5XFAD brain. In 5XFAD neuronal soma, intraneuronal Aβ42 accumulates in puncta that co-label for Transferrin receptor and LAMP-1, indicating endosomal and lysosomal localization, respectively. In addition, in young 5XFAD brains, we observed activated Caspase-3 in the soma and proximal dendrites of intraneuronal Aβ42-labeled neurons. In older 5XFAD brains, we found activated Caspase-3-positive punctate accumulations that co-localize with the neuronal marker class III β-tubulin, suggesting neuron loss by apoptosis.
Together, our results indicate a temporal sequence of intraneuronal Aβ42 accumulation, Caspase-3 activation, and neuron loss that implies a potential apoptotic mechanism of neuron death in the 5XFAD mouse.
KeywordsIntraneuronal Aβ42 5XFAD Alzheimer’s disease Amyloid-β Caspase-3 Neuron loss Apoptosis
5 familial AD mutations
β-site APP cleaving enzyme-1
Lysosomal-associated membrane protein 1
Amyloid precursor protein
Bcl-2-associated X protein
The histopathology of Alzheimer’s disease (AD) is characterized by two hallmark lesions, extracellular amyloid-β plaques made of the Aβ peptide, and intracellular neurofibrillary tangles composed of hyperphosphorylated tau (reviewed in [1–5]). In addition to the presence of plaques and tangles in the brain, considerable neuron loss is also a cardinal feature of AD, but the mechanisms of neural cell death are unclear. Importantly, familial AD mutations (FAD) in the genes for amyloid precursor protein (APP), presenilin 1 (PS1), and presenilin 2 (PS2) that cause AD implicate Aβ as an initiating factor in AD pathogenesis (reviewed in [5, 6]). These FAD mutations increase the production of Aβ42, the 42-amino acid form of the peptide, from APP, which is sequentially cleaved by the β- and γ-secretase enzymes to release the peptide. These results, among others, strongly suggest that Aβ42 plays a central early role in the pathophysiology of AD that ultimately leads to the neuron loss and dementia observed in the disorder.
The mechanism by which Aβ42 exerts neurotoxicity is poorly understood; many mechanisms have been hypothesized, but none have been definitively proven. The accumulation of intraneuronal Aβ42 has been observed in the brains of AD patients and APP transgenic mice, and studies suggest that intraneuronal Aβ42 plays a role in neurodegenerative processes relevant to AD (reviewed in [7–9]). Frank neuron loss has been observed in two aggressive amyloid plaque transgenic mouse models that also exhibit accumulation of intraneuronal Aβ42 prior to plaque formation: the 5XFAD and APPSLPS1K1 lines [10, 11]. These transgenic models express multiple FAD mutations that additively increase Aβ42 production. In the case of the 5XFAD model, the mouse overexpresses APP with K670N/M671L (Swedish mutation ), I716V (Florida mutation ), and V717I (London mutation ), and PS1 with M146L and L286V mutations . Individually, each FAD mutation enhances Aβ42 generation, but together they act synergistically in the transgenic mouse to predominantly make Aβ42. Consequently, 5XFAD mice represent a very aggressive amyloid deposition model that develops intraneuronal Aβ42 at 1.5 months, plaques at 2 months, memory deficits at 4 months, and neuron loss at 9 months of age . These characteristics make 5XFAD mice a robust model for investigating the role of intraneuronal Aβ42 in neuron loss.
Here, we have examined the process of neuronal death in 5XFAD mice and found a correlation between intraneuronal Aβ42, neuron loss, and Caspase 3 activation in large pyramidal neurons of the brain. These results suggest a potential role for an apoptotic mechanism in intraneuronal Aβ42-mediated neuron loss, and may have relevance for neuronal death in AD.
5XFAD mice exhibit progressive neuron loss in cortical Layer 5 and subiculum
Together, our results confirm and extend our previous qualitative work  and the quantitative study of Jawhar and colleagues  demonstrating neuron loss in the 5XFAD model. In addition, our data indicate that 5XFAD neuron loss occurs at earlier ages than previously shown and appears to begin when mice are about 6 months old. Thus, the 5XFAD model should prove useful for studies of Aβ-associated neurodegeneration and neuron loss.
5XFAD mice exhibit intraneuronal Aβ42accumulation that precedes amyloid plaque formation
5XFAD mice develop intraneuronal Aβ42at an early age
Previous studies have suggested that intraneuronal Aβ is a potentially important factor participating in AD pathogenesis (reviewed in [8, 21, 22]). Our initial work demonstrated the presence of intraneuronal Aβ42 in 5XFAD mice starting at ~1.5 months , an observation that has been recently confirmed independently by several groups [10, 16, 23–26]. To verify intraneuronal Aβ42 in 5XFAD mice, we immunostained brain sections from 1.5, 2, and 4 month old mice with anti-Aβ42 selective antibody (Figure 3G-J). At 1.5 months, before the appearance of amyloid plaques, many large pyramidal neurons in cortical Layer 5 displayed numerous strongly stained Aβ42-positive puncta within their soma (Figure 3G). Intraneuronal Aβ42 staining was present when amyloid deposition began at ~2 months and persisted as plaque number increased with age (Figure 3H,I arrows). However, intraneuronal Aβ42 became more difficult to visualize over time, suggesting that either intraneuronal Aβ42 accumulation is transient , or intracellular Aβ42 signal is obscured as plaque staining intensifies with age. Non-transgenic littermate control sections lacked both plaques as well as intraneuronal Aβ42-positive puncta (Figure 3J), as expected.
The presence of intraneuronal Aβ in AD and APP transgenic brains has been controversial. One criticism against intraneuronal Aβ has been that anti-Aβ antibodies also recognize full-length APP and APP-CTFs, depending on the antibody (e.g., see ). To ensure that our selective anti-Aβ42 antibody was detecting bona fide intraneuronal Aβ42 and was not cross-reacting with APP or APP CTFs, we performed antibody titration experiments on both 5XFAD; BACE1+/+ and 5XFAD; BACE1−/− brain sections. At a primary antibody dilution of 1:20,000, punctate Aβ42-positive staining was detected in the soma of large pyramidal neurons in cortical Layer 5 of 5XFAD; BACE1+/+ sections (Figure 3K-O), but no signal was observed in 5XFAD; BACE1−/− sections (Figure 3P-T). Because 5XFAD; BACE1−/− mice express the APP transgene but do not generate Aβ , these results strongly support the conclusion that our anti-Aβ42-selective antibody does indeed detect intraneuronal Aβ42 rather than full-length APP or APP CTFs.
To further confirm that our anti-Aβ42-selective antibody does not cross-react with APP, we performed confocal immunofluorescence microscopy on 5XFAD; BACE1+/+ and 5XFAD; BACE1−/− brain sections co-stained with anti-Aβ42-selective and APP antibodies. 5XFAD sections showed punctate labeling of Aβ42 throughout the neuron soma (Figure 4A,B,E,F,I), while anti-human APP (6E10) staining revealed additional punctate and cell-surface labeling that were negative for Aβ42 (Figure 4B). As expected, Aβ42 signal was completely absent in the 5XFAD; BACE1−/− sections (Figure 4C,D), but transgenic APP labeling was evident (Figure 4D), again demonstrating that our Aβ42-selective antibody does not recognize full-length APP or APP CTFs at the concentration used. Co-staining 5XFAD brain sections with anti-Aβ42 and a second anti-APP antibody (Karen, ) produced similar results (Figure 4E). Together, our data demonstrate that the anti-Aβ42 C-Terminal neo-epitope antibody does not have significant cross-reactivity with APP, and that punctate Aβ42 immunoreactivity resides within 5XFAD neuronal soma.
Intraneuronal Aβ42co-localizes with endosomal and lysosomal markers
The punctate Aβ42 immunolabeling in neuronal soma suggested that Aβ42 might be contained within membrane-bound subcellular compartments. Previous studies have indicated the presence of Aβ42 in endosomes and late endosomal multivesicular bodies in other models [30–32] and in cathepsin-D-positive puncta of 5XFAD mice . To further investigate the subcellular localization of Aβ42 in 5XFAD neurons, brain sections were co-immunostained with anti-Aβ42 and anti-Transferrin receptor or anti-LAMP1 antibodies to identify endosomes or lysosomes, respectively, and imaged by immunofluorescence confocal microscopy. Indeed, we observed partial co-localization of punctate Aβ42 staining with signals from both Transferrin receptor and LAMP1 (Figure 4F-K). Although Aβ42 co-localization was observed with both markers, neither was exclusive. These results suggest the accumulation of intraneuronal Aβ42 throughout the endosomal-lysosomal system.
5XFAD brains exhibit increased Caspase-3 activation
Despite years of study, the mechanisms of neuron loss in AD are still incompletely understood. Apoptosis has been proposed as a potential mechanism leading to the neuronal death observed in this devastating neurodegenerative disorder (reviewed in [34–36]). Here, in the 5XFAD amyloid plaque mouse model, we have demonstrated progressive loss of neurons that correlates strongly with both intraneuronal Aβ42 and Caspase-3 activation. We observe a time course in 5XFAD mice in which intraneuronal Aβ42 first appears at 1.5 months of age in large pyramidal neurons of Layer 5 cortex and subiculum, concomitant with the observation of activated Caspase-3 in these neurons, but prior to amyloid deposition. By 4 months of age, 5XFAD mice exhibit amyloid plaques and the appearance of Caspase-3 positive puncta in Layer 5 cortex and subiculum that suggest neurodegeneration, but at a time when statistically significant neuron loss is not yet evident. At 6 months of age, 5XFAD mice as a whole show a trend toward neuron loss, in which Layer 5 neurons appear to split into two subgroups, those that display neuron loss and those that do not. However, by 9 months, neuron loss is statistically significant in Layer 5 cortex of 5XFAD mice. This time course, although correlative, suggests the possibility that intraneuronal Aβ42 triggers Caspase-3 activation, which in turn induces apoptosis-mediated neurodegeneration and eventual neuron loss in 5XFAD mice.
The 5XFAD mouse, along with the APPSLPS1KI, is one of a few known mouse models that exhibits significant neuron loss in addition to displaying other AD hallmarks such as amyloid plaques [10, 11, 16]. Our unbiased stereological counting of 5XFAD mice revealed a significant decline in large pyramidal neurons of the Layer 5 cortex starting at 9 months of age. Our data is consistent with previous studies, except we have demonstrated that significant 5XFAD neuron loss begins at an earlier age than previously reported . Interestingly, a division develops at 6 months that appears to separate 5XFAD mice into those that exhibit significant neuron loss and those that have yet to surpass the point where neuron loss is occurring at a measurable rate. The ~6 month age period marks the first appearance of oligomeric N-terminally truncated Aβ peptides with pyroglutamate and an increasing decline of motor abilities in 5XFAD mice [16, 23].
Consistent with previous studies [10, 32], we observed intraneuronal Aβ42 as small puncta located in neuronal soma of 5XFAD mice. The intraneuronal Aβ42 became progressively more difficult to detect in 5XFAD sections after 1.5 months of age, presumably because highly concentrated Aβ42 plaques effectively blocked visualization of smaller and less prominent intraneuronal Aβ42 puncta. Despite this difficulty, 5XFAD mice exhibited a subset of neurons that harbored intraneuronal Aβ42 past 4 months. The early appearance of intraneuronal Aβ42 and its correlation with neuron loss in 5XFAD mice is also shared with the APPSLPS1KI transgenic mouse model [11, 33, 37, 38].
We found that intraneuronal Aβ42 in 5XFAD mice was partially co-localized to endosomes and lysosomes, as indicated by co-labeling for Transferrin receptor and LAMP1, respectively. This corresponded well with previous findings that intraneuronal Aβ42 localizes to late endosomal multivesicular bodies  or lysosomes ; the latter observation is particularly significant, as it was obtained in the 5XFAD mouse. Intraneuronal Aβ42 localization to endosomes and lysosomes is also supported by previous studies that have proposed a mechanism in which the ubiquitin-proteasome system is disrupted, causing impaired protein degradation and increased Aβ accumulation [39, 40].
It has been hypothesized that apoptosis is involved in the neuron loss seen in AD (reviewed in [34–36]). For example, activated Caspase-3 immunoreactivity has been reported to be present in AD brain [41, 42] and in APP transgenic mice [43, 44]. Interestingly, Aβ42 has been shown to cause cytochrome c release from mitochondria , which activates Caspase-3 and induces apoptosis, thus providing a potential mechanism for intraneuronal Aβ42-mediated neuron loss in 5XFAD mice.
In 5XFAD mice, activated Caspase-3-positive puncta were present throughout Layer 5 cortex and subiculum, often in close proximity to Aβ42-positive plaques. Caspase-3 activation was not apparent in both aged matched non-transgenic littermate and 5XFAD; BACE1−/− mice. We speculate that these activated Caspase-3-positive puncta are likely the remnants of neurons that died by apoptosis, a conclusion supported by the observation of co-localized activated Caspase-3 and neuron-specific class III β-tubulin (Figure 6F). The Caspase-3 puncta in 5XFAD brain are reminiscent of activated Caspase-3-positive AD neurons that exhibit typical features of apoptosis such as condensed cytoplasm and shrunken nuclei [46, 47]. In addition, the association of activated Caspase-3 and neuron loss in 5XFAD mice is consistent with the findings of apoptotic neuron death and “apoptotic bodies” found in a different PS/APP (PS1M146L and K670N/M671L) mouse model of AD . While not always coexistent, we observed large pyramidal neurons from 1.5 month-old 5XFAD mice that contained both punctate intraneuronal Aβ42 within the neuron soma and activated Caspase-3 throughout the cell body and proximal dendrites (Figure 7K, L). We speculate that these neurons could represent an early stage of apoptotic neuronal degeneration and death. Interestingly, a recent study found activated caspase-3 in dendritic spines of hippocampal neurons in the Tg2576 APPswe transgenic mouse model , supporting the dendritic localization of activated Caspase-3 that we have observed in 5XFAD mice.
The 5XFAD mouse has proven to be a useful model of amyloid deposition in the brain. Here, we have extended the utility of this transgenic mouse to the investigation of neuron loss that may have relevance to AD. We present compelling correlative data that associates intraneuronal Aβ42 with Caspase-3 activation and neuron death. Our data suggest a mechanism whereby intraneuronal Aβ42 accumulation in the endosomal-lysosomal pathway leads to activation of Caspase-3 and apoptotic neuron death in 5XFAD mice. Although some caution is warranted in interpreting results from an aggressive amyloid mouse model that overexpresses multiple FAD mutations, our results are consistent with evidence of activated Caspase-3 and apoptotic neuron death in human AD. Our study provides a framework that, with further elucidation of mechanisms, could lead to a deeper understanding of neuron loss in AD.
5XFAD transgenic mice
5XFAD mice overexpress the K670N/M671L (Swedish), I716V (Florida), and V717I (London) mutations in human APP (695), as well as M146L and L286V mutations in human PS1. The generation of the 5XFAD mice has been described previously .
Mice were euthanized following IACUC approved procedures to minimize pain. After dissection, one hemibrain of each mouse was fixed in paraformaldehyde/PBS and cryo-preserved in 30% sucrose/PBS containing 0.01% sodium azide. Brains were sectioned sagittally on a freezing microtome at 30 μm. Serial sections were collected in 0.1 M phosphate buffer, pH 7.6.
Every sixth section was taken in sequential order, thoroughly washed in PBS, and then mounted on charged slides. Sections were allowed to dry between 1.5 to 2.5 hours. Slides were then taken through the cresyl violet staining procedure, where at each step the slides were slightly agitated every 15 seconds to ensure uniform staining. The sections were first briefly rinsed in water and then stained in 0.5% cresyl violet for 10 min. Sections were then rinsed in water briefly and then dehydrated in 50%, 70%, and 90% EtOH solutions for 30 seconds each. Sections were then de-stained in a 95% EtOH/1.4% glacial acetic acid solution for 1 min. Finally sections were dehydrated in a series of alcohols, cleared in xylene, and cover-slipped. The Layer 5 of the cortex was sampled for each section (1600 μm was examined starting at the point dorsal to the anterior most part of CA3 and then moving to the posterior of the brain). A Nikon (Tokyo, Japan) Eclipse 80i microscope with a motorized stage, StereoInvestigator 9 (Microbrightfield, Williston, VT, USA), and a 60x air lens was used to perform the stereological counting. Neuronal nuclei were then randomly sampled from the defined the Layer 5 using optical dissector probes, and a neuronal population was determined using a 2 μm top and bottom guard zone. Comparison of means using the t-test was performed using GraphPad Prism (GraphPad Software, Inc., San Diego, CA).
Free-floating sections were washed in PBS with 0.4% Triton and then blocked in 3% goat serum for one hour. Sections were then incubated at 4°C on a shaker overnight with the following primary antibodies: rabbit monoclonal anti-Aβ42 (1:1000; Invitrogen; 700254) or rabbit monoclonal anti-cleaved Caspase-3 (1:1000; Cell Signaling; 9664). After more washes, the sections were incubated overnight at 4°C on a shaker with secondary biotinylated goat anti-rabbit at 1:1000. After a third set of washes, The Vector Laboratories (Burlingame, CA) ABC kit was used with 3,3′-Diaminobenzidine (DAB) as chromogen to visualize the reaction product. The sections were then mounted on charged slides, allowed to dry, dehydrated in a series of alcohols, cleared in xylene, and cover-slipped. Light microscopy was performed on a Nikon (Tokyo, Japan) Eclipse 80i microscope.
Immunofluorescence confocal microscopy
Free-floating sections were treated with TBS with 0.1% SDS and 5 mM DTT for 30 min on a shaker. Sections were washed in TBS and then blocked in 5% goat or donkey serum for one hour (respective to the secondary being used). Sections were incubated at 4°C on a shaker overnight with the following primary antibodies: rabbit monoclonal anti-Aβ42 (1:1000; Invitrogen; 700254), mouse monoclonal anti-Aβ42 (1:1000; Covance; SIG-39142), rabbit monoclonal anti-cleaved Caspse-3 (1:1000; Cell Signaling; #9664), rat monoclonal anti-Transferrin receptor (1:500; Abcam; ab60344), rat monoclonal anti-LAMP1 (1:500; Abcam; ab25245), 6E10 mouse monoclonal anti-Aβ (1:1000; Chemicon; MAB1560), Karen goat polyclonal anti-N-terminal APP (1:500; ), DAPI (Invitrogen). After more washes, the sections were incubated for 1.5 h at room temperature on a shaker with Invitrogen Alexa Fluor secondary antibodies at 1:1000 (Donkey anti-Mouse or Rabbit-488 or 594; Goat anti-Mouse or Rabbit-488 or 594) and DAPI at 300nM. The sections were then mounted on charged slides and immediately cover-slipped using ProLong Gold (Invitrogen). Confocal images were captured on a Nikon (Tokyo, Japan) Eclipse C1si spectral laser scanning confocal microscope.
We would like to thank Sarah Cole for intellectual contributions and performing immunohistochemistry, Erika Maus and Patrick Earley for assistance with mouse work, Virginia Lee and Lester Binder for the kind gifts of Karen and TUJ1 antibodies, respectively, and Lester Binder, Richard Miller, and Changiz Geula for helpful comments and advice about this project. This work was supported by funding from the Alzheimer’s Association, MetLife Foundation, and NIH (R01 AG022560). Confocal imaging work was performed at the Northwestern University Cell Imaging Facility generously supported by NCI CCSG P30 CA060553 awarded to the Robert H Lurie Comprehensive Cancer Center.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.