Background
The neuropathological hallmarks of Alzheimer’s disease (AD) include amyloid-β (Aβ) plaque accumulation, neurofibrillary tangle formation, and synaptic and neuronal loss. How these factors ultimately contribute to memory loss and cognitive deficits that clinically characterize the disease remains unclear. In AD brains, Aβ can spontaneously aggregate and thereby assume a variety of conformational states ranging from monomers to soluble oligomers, protofibrils, and insoluble fibrils, which assemble to form extracellular plaques [
1,
2]. Further characterization of this array of Aβ conformations has proven critical for deciphering which of these may be neurotoxic. The insoluble fibrillar Aβ (fAβ) found in the extracellular plaques in post-mortem AD brains has long been postulated as the initiating agent in the neurodegenerative cascade of the disease. However, many recent studies have demonstrated that the amount and distribution of amyloid plaques does not correlate with synaptic or neuronal loss, and the onset of cognitive decline in AD [
3‐
9]. In contrast, soluble oligomeric Aβ assemblies (oAβ, also called Aβ-derived diffusible ligands), which range in size from tetramers to dodecamers and can accumulate intracellularly, show correlation with synaptic loss and synaptic impairment in many
in vitro studies. Soluble oAβ binds preferentially to synapses [
10], and addition of oAβ to mouse hippocampal slices results in inhibition of long-term potentiation [
11]. This correlation has also been observed
in vivo; the injection of oAβ directly into the hippocampus of rats resulted in deficits in learning and memory [
2,
12,
13]. While these data suggest that soluble oAβ represents the neurotoxic species in AD over insoluble fibrillar forms, the relationships between oAβ, neurodegeneration, and cognitive decline remain poorly defined, with most studies having only examined the toxicity of oAβ
in vitro or
in vivo in mouse models that possess soluble oAβ as well insoluble fAβ, and Aβ plaques (see [
13,
14] for review). Data generated from these mouse models produced results that can be difficult to interpret due to the presence of multiple Aβ conformations.
Mutations in amyloid precursor protein (
APP) are observed in patients with familial early-onset AD or dementia caused by cerebral amyloid angiopathy. Of the
APP mutations causing AD or cerebral amyloid angiopathy, four occur at the E693 position of the protein, the Dutch (E693Q) [
15], Arctic (E693G) [
16] and Italian (E693K) [
17] mutations and a deletion (E693Δ) [
18]. In contrast to the pathological amyloid deposition observed in AD, patients who carry the
APP E693G (Arctic) or E693Δ variants show little or no fibrillar Aβ as detected by amyloid imaging [
18,
19]. Current imaging technologies cannot detect soluble oAβ, which may be present in the brain, affect synaptic function and lead to the cognitive deficits observed in these patients. The present investigation sought to examine the effects of soluble oAβ on neuronal and synaptic structure in the APP E693Q (“Dutch”; DU) mouse model of AD that displays intraneuronal accumulation of soluble oAβ with no detectable plaques. This mouse model expresses the E693Q mutation of the
APP [
20]. Missense mutations in
APP located near the middle of the Aβ domain influence the propensity of Aβ to form oligomeric assemblies by disrupting the salt bridges on the protein that typically stabilize parallel β-pleated sheets and favor fibril and plaque formation, thereby promoting the formation and intraneuronal accumulation of oAβ [
21]. Severe meningocortical vascular deposition of Aβ in patients with hereditary cerebral hemorrhage with amyloidosis caused by the DU mutation has also been described. Interestingly, these patients consistently develop cerebral hemorrhages but rarely display significant parenchymal amyloid plaque accumulation [
22,
23]. This was initially proposed to be related to the ratio of Aβ42/40, with Aβ40 being the dominant species, however subsequent studies revealed Aβ42 also plays a role in vascular amyloid formation [
22,
23]. Recent work by Gandy and colleagues indicates that the level of soluble oAβ in the DU mouse model correlates with diminished performance in the water maze compared to non-transgenic wild type (WT) littermates at 12 months [
20], indicating that DU mice, which do not demonstrate extracellular deposits, exhibit perturbed hippocampus-associated spatial learning and memory.
Our goal was to complement these behavioral findings by performing quantitative analyses of neuronal pathology, including dendrite morphology, as well as spine and synapse numbers in individual hippocampal CA1 neurons. These morphological features of the neuron are the site of critical memory forming processes [
24‐
26]. Twelve month-old DU mice were assessed, along with non-transgenic WT littermates, for comparison of dendritic length and complexity, and quantification of dendritic spines and synapse densities. Structural components of synapses such as postsynaptic density (PSD) length and synapse head diameter of individual pyramidal CA1 neurons were also examined. We show that DU mice displayed morphological changes in CA1 neurons compared to WT mice, including significantly reduced dendritic arborization of apical dendrites along with decreases in PSD length of synapses on mushroom spines. These structural alterations to neurons in DU mice support the concepts that the soluble oAβ species have adverse effects at the synaptic level along with major structural disturbances, and that changes at the synapse correlate with early cognitive impairments in AD.
Discussion
Aβ and its proposed neurotoxicity has been the focus of AD research since it was identified in the brains of AD patients three decades ago [
30,
31]. The discovery that Aβ has a variety of conformations, and the lack of correlation between cognitive decline and plaque accumulation, has prompted researchers to reassess the ‘amyloid cascade hypothesis’ that dominated early AD research [
32]. Recent advances in early AD diagnosis have enabled amyloid PET imaging, using ligands such as florbetapir [
33] and Pittsburgh Compound-B [
34] to detect Aβ deposits. However, these studies have also validated that hippocampal neurodegeneration is unrelated to rate of Aβ deposition [
35]. The strong correlation between soluble conformations of Aβ and cognitive impairments, along with synaptic dysfunction and loss has shifted the focus of amyloid research towards soluble, oligomeric Aβ conformations as the ‘toxic species’ [
32]. Indeed, monomeric and oligomeric forms of Aβ have been reported to accumulate in synapses of AD brains as compared to controls [
36]. However, new tools and models to discriminate between different Aβ aggregation states have not been available until recently [
37], and despite reports of altered dendritic spine and synaptic morphology in a variety of AD experimental models, there is limited knowledge about how soluble oAβ may affect these structural parameters in a model that does not also accumulate amyloid plaques.
In the current study, we have utilized a mouse model of AD that expresses the Dutch variant of APP.
In silico modeling of Aβ fibrillization [
38], cell-free aggregation experiments [
39] and
in vitro experiments [
40,
41] showed that the Dutch mutation increased fibrillization of Aβ. Indeed, the Dutch mutation increases deposition of Aβ in vascular tissue but has the opposite effect in parenchyma of the transgenic mice, as is seen in patients carrying this mutation [
42]. In contrast to the mice used in the Herzig study, which expressed both normal and Dutch APP, the DU mice we used in the current study expressed Dutch APP and only produces soluble oAβ, with no detectable plaques up to 30 months of age [
20] (Figure
1). The DU mice display cognitive deficits in spatial memory tasks at 12 months of age, indicating that soluble oAβ may be inducing these effects in the absence of fAβ. Here we demonstrate that CA1 pyramidal neurons exhibit changes to neuronal morphology and total spine and synapse density. We selected the CA1 region for analysis since it is implicated in long-term contextual memory retrieval [
43,
44], which is affected in AD [
45]. Neocortical pyramidal neurons possess extensive apical and basal dendritic trees, which integrate information from thousands of excitatory and inhibitory synaptic inputs [
6]. The morphological features of a dendrite, such as its length and complexity, can influence how a neuron processes and transmits information [
46‐
49], and many
in vivo studies do report neuronal atrophy with proximity to fAβ in the brain and subsequent functional deficits (see [
6] for review). The effects of oAβ are less well characterized, although oAβ has been implicated in alterations to overall dendritic spine number and type in AD. Most studies on spines in AD models report either a spine loss, or a shift in the proportion of thin, stubby, or mushroom spine types, which are proposed to have unique roles, complementing their distinctive morphology [
3,
50‐
52]. For example, thin spines contain a predominantly greater number of the
N-methyl-D-aspartate glutamate receptors (NMDARs) compared to large, mushroom spines that contain more 2-amino-3-(3-hydroxy-5-methyl-isoxazol-4-yl)propanoic acid receptors (AMPARs), making synapses on mushroom spines functionally stronger [
28]. This suggests that the more plastic, thin spines are linked to learning, whereas mushroom spines may represent more stable, ‘memory’ spines [
28,
29]. A growing number of reports support the notion that spines and synapses are the initial site of Aβ-mediated neurotoxicity [
24‐
26,
53], and it is probable that more dramatic oAβ-driven disturbances in DU mice are occurring at the synaptic receptor level, which we did not examine here. Excitatory glutamatergic NMDARs, AMPARs and other critical synaptic proteins may be affected in DU mice, with reduced expression and altered distribution of NMDA and AMPA receptors. Other
in vitro and
ex vivo experimental models have reported direct binding of soluble Aβ to synaptic proteins at the PSD [
53‐
55] and these observations are supported by the decrease in PSD length in DU mice we observed in this study.
The significant decrease we observed in PSD length of synapses only on apical mushroom spines may reflect disturbances in synaptic function and memory impairment in the presence of oAβ. A similar finding was reported by Nicholson et al. (2004) in their study of aged, learning-impaired Long Evans rats. These authors reported a significant (~30%) decrease in PSD length of CA1 SR synapses in the impaired group compared to control rats [
56], and noted that the decrease was only apparent in perforated synapses, which are defined by a discontinuity of the electron dense plate on the post-synaptic membrane into two or more parts [
57,
58]. Although the decrease in PSD length we observed in DU mice for synapses in the same brain region were not as dramatic, our data is supported by this study as perforated synapses typically occur only on mushroom spines implicated in memory formation [
28,
29]. This is because the shape of the PSD can increase in complexity with increasing spine size [
57]. Furthermore, perforated synapses have been proposed to enhance synaptic efficacy by increasing the surface area for receptor activation [
56,
59,
60]. A decrease in PSD length would therefore reduce glutamate receptor content and impair synaptic efficacy, leading to functional disturbances [
60]. Within the context of the current study, a decrease in PSD length suggests that hippocampal perforated synapses are selectively affected by soluble oAβ, correlating with the observed cognitive deficits in 12 month-old DU mice. Alternatively, it is possible that although we did not observe any changes in perforated and non-perforated synapses or total synapse number, the DU mice may have an increased proportion of AMPA-type ‘silent’ synapses. These express NMDARs, but lack functional AMPARs [
61], and do not generate a synaptic response following glutamate release at normal resting membrane potentials [
62]. In electron micrographs, they are structurally identical to functional synapses, and can only be distinguished by a lack of AMPA immunoreactivity [
62‐
67]. When a study by Geinisman et al. (2004) found no changes in perforated and non-perforated synapses, and total synapse numbers in the SR dendritic domain of CA1 neurons of cognitively impaired, aged rats compared to control rats, they argued that this may be due to an increased proportion of silent synapses [
62]. Because silent synapses may become functional during postnatal development and the induction of hippocampal long-term potentiation, it is also possible that functional synapses in CA1 pyramidal neurons become increasingly silent with age [
62], leading to a loss of synaptic activity without diminished synapse numbers. Further support for an oAβ-driven increase in silent synapses comes from other work using both the transgenic mice carrying the Swedish APP mutation and the external application of oAβ to rat cortical neurons, which found that oAβ altered the distribution and reduced the expression of Ca
2+/calmodulin-dependent protein kinase II (CaMKII) [
68]. CaMKII is a critical scaffolding protein highly enriched at PSDs that is required for transporting AMPARs to the synapse, and switching synapses from silent to functional through AMPA delivery [
69,
70]. A reduction in the expression of CaMKII following oAβ exposure therefore provides a mechanism for increased silent synapses to occur. Interestingly, no changes in dendritic spine density were also found by another group using oAβ externally applied to rat cortical cultures compared to controls [
68]. Further analysis of the DU mice using immunogold electron microscopy (EM) will help to clarify if soluble oAβ accumulated
in situ influences the proportion of silent synapses or dramatically alters the expression and distribution of synaptic receptors in the CA1. Additional behavioral and immunogold EM analyses in older DU mice (18 and 24 month-old) will also reveal whether their cognitive deficit and neuropathology worsens with age.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
Study concept and design: KAP, DLD. Acquisition of data: KAP, AS, FY, HB, DLD. Analysis and interpretation of the data: KAP, DLD. Drafting of the manuscript: KAP, MV, DLD. Statistical analysis: KAP. Critical revision of the manuscript for intellectual content: KAP, MV, MEE, DLD. Material support: MEE. All authors read and approved the final manuscript.