Introduction
Alzheimer’s disease (AD) is a devastating progressive neurodegenerative disorder, mainly characterized by the accumulation of amyloid-beta (Aβ) plaques and neurofibrillary tangles (NFTs), which leads to dementia [
40]. Most of AD patients develop the late-onset sporadic form of AD (sAD), while the early-onset familial form of AD (fAD) is rare (< 1% of AD cases) [
17,
28,
32]. Ageing has been identified as the greatest known risk factor of sAD, with the prevalence doubling every 5 years for people over 65 and reaching nearly one third by the age of 85 [
40]. As the ageing population is increasing, the number of AD patients suffering from dementia will raise considerably [
26].
Evidence from studies on fAD (caused by genetic mutations) supported the idea, referred as “the amyloid hypothesis”, that the abnormal accumulation of extracellular soluble Aβ (sAβ) peptides is initiating AD pathology [
30,
31]. Under normal conditions, sAβ peptides are naturally generated from the proteolysis of the amyloid precursor protein (APP), which is a single-pass transmembrane protein expressed at high levels in the brain and concentrated in the synapses of neurons [
38]. Further, sAβ monomers are thought to have a physiological role in controlling synaptic activity, excitability and cell survival [
39]. Under pathological conditions and upon ageing, a metabolic dysregulation seems to cause the accumulation of sAβ peptides in the extracellular space, which oligomerize and aggregate, forming insoluble Aβ plaques [
16,
57]. It is believed that sAβ oligomers are more synaptotoxic than the plaques, altering synaptic transmission and causing synapse loss and neuronal death [
12,
32,
35], mainly in the cerebral cortex and certain subcortical regions [
2], resulting in a progressive loss of cognitive functions [
16].
Despite the important efforts to develop therapeutics, there is currently no cure available for AD. As the first symptoms of AD appear decades after the onset of the disease, scientists have moved their focus on the early stage of the disease, including both therapeutic treatments and early diagnosis. With the evolution of neuroimaging techniques, some functional AD-related alterations have been highlighted, particularly using resting-state functional magnetic resonance imaging (rsfMRI). This in vivo non-invasive imaging technique measures the blood oxygen level-dependent (BOLD) low-frequency signal fluctuations of neuronal activity to indirectly evaluate brain network functions [
7,
15]. The temporal correlation of these fluctuations between spatially distinct brain regions has been suggested to reflect the brain’s functional connectivity (FC). Therefore, using rsfMRI, we can detect alterations at the neuronal network level, reflecting impaired communication between brain regions caused by synaptic dysfunction. In addition, it has been demonstrated that AD patients and individuals with mild cognitive impairment displayed disturbed FC in specific brain networks, such as the default mode network (DMN) [
3,
6,
23,
24]. The DMN was identified as a network of anatomically distant brain regions which show highly correlated synaptic activity when a person is at rest [
10,
22]. Thus, rsfMRI performed at early stages can provide important information on the FC and its potential disturbances, crucial for early diagnosis of AD.
Increasing evidence towards oligomeric sAβ as an early biomarker, instead of plaques that appear later in the course of disease pathology, made the research focus on the role of sAβ oligomers in AD development. For instance, synaptic transmission and plasticity were altered months before plaque deposition in the hippocampus of young AD transgenic mice [
33]. Behavioural deficits were also found in the pre-plaque stage [
27]. In 2012, Busche et al. showed, using in vivo two-photon Ca2+ imaging, that sAβ oligomers caused very early functional impairment in AD transgenic mice due to hippocampal hyperactivity [
13]. Moreover, they also reported that sAβ oligomers altered the slow-wave propagation, causing a long-range slow-wave activity breakdown [
14]. Interestingly, rsfMRI performed in different transgenic mouse models of AD initially reported decreased FC at a stage of abundant plaque deposition [
21,
44]. As the research focus moved from insoluble amyloid deposits towards oligomeric sAβ, rsfMRI performed before plaque deposition revealed an early pathological hypersynchronisation of brain resting-state networks, especially in the DMN [
45]. This evidence of oligomeric sAβ-induced synaptotoxicity has, however, come directly from transgenic developmental-onset mouse models of amyloidosis, i.e. the overexpression of APP starts from embryonic or very early postnatal developmental stage. As overexpression, and consequently, overproduction of APP and Aβ, occurs during brain development, this might affect the cell signalling and synapse formation, causing artificial phenotypes unrelated to AD pathology [
28,
37]. Moreover, immature and mature mice might respond differently to the high exposure of sAβ species, making the neuronal circuits more sensitive or resilient to sAβ [
51]. Therefore, it is crucial to evaluate the toxicity of sAβ oligomers on mature mice. However, all available transgenic models of amyloidosis contain the fAD mutations, thus modelling the amyloid pathology of fAD (developmental-onset) and not sAD [
28].
In order to evaluate the impact of oligomeric sAβ accumulation on the brain neuronal networks in mature mice, we used the inducible Tet-Off APP mouse model of amyloidosis (line 107) in which the APP overexpression can be controlled via a specific diet [
29]. APP overexpression was switched-on when the mice were 3 m of age, when the brain can be considered as mature [
25]. As the most drastic developmental changes in both myelination and cortical thickness have already taken by that age, this model was referred as mature-onset APP expression, which is in a context resembling closer the human sAD amyloid pathology. Using rsfMRI and ex vivo brain analyses, we investigated the toxicity of increased sAβ oligomers in mature-onset Tet-Off APP mice on brain neuronal networks.
Discussion
This study aimed to address the effects of sAβ on the neuronal networks in mature-onset amyloidosis TG mice. By inducing APP overexpression only once the main developmental processes have occurred, we avoided the possible impact of protein overexpression on the brain during postnatal development and its potential consequences later on, once the brain is mature. We conjecture, that this model can thus provide valuable information on the role of amyloid pathology in sAD, for which this cascade is usually triggered well post brain development. Even though the TG mice are bearing mutations of fAD, the clear advantage of this model over other developmental-onset AD mouse models is the possibility for controlling transgene APP expression in time. Despite this clear advantage, one should keep in mind that administration of DOX could possibly alter and inhibit microglia activation, which can contribute in the mice’s phenotype. Against this hypothesis, however, the TG mice developed the amyloid pathology, with microglia responses similar to other models [
29]. While only resembling amyloid aspects observed in human AD pathology [
17,
28], the inducible Tet-Off APP mouse model used here is highly relevant for studying Aβ-induced synaptic dysfunction on the neuronal networks in mature-onset TG mice. A different mouse line of the Tet-Off APP model was used by Sri and colleagues in 2019 and found changes in the behavioural phenotype and electrophysiological measures indicating important differences between developmental and mature-onset APP expression [
51].
Here, by controlling chimeric APP expression in the Tet-Off APP model, we could for the first time demonstrate spatiotemporal dynamics of brain networks in mature-onset mice by means of non-invasive and translational rsfMRI. Particularly, we could assess the impact of increased sAβ levels on the brain networks from the initial stages up to abundant Aβ plaque deposition, and correlate the neuronal network alterations to sAβ-induced synaptic toxicity.
Of note, at week 0 post DOX treatment, the TG mice did not show any significant difference in FC compared to the controls. While the TG animals carry genetic mutations, those manipulations had no detectable or apparent effect on the normal brain networks function during the suppression phase of APP transgene expression. However, 8w later, the TG animals showed a significant hypersynchronisation in the DMN, in which both intra- and inter- region FC were significantly higher compared to the controls (Fig.
3b,
4b,
5 and Additional file
2: Figure S2). This hypersynchronisation of the neuronal networks in the TG mice coincided with a significant increase of total sAβ levels in absence of plaques (Fig.
6), suggesting a sAβ-induced toxicity affecting the brain networks and the communication between them before plaque deposition. Moreover, this neuronal network hypersynchronisation was not associated with any detectable inflammatory responses (Fig.
7), while neuroinflammation is an important feature of human AD pathology. Thus, our current results showed an early hypersynchronisation of brain networks associated with sAβ increases similarly to previously reported studies using developmental-onset mouse models [
45]. Noteworthy, the brain networks affected were strongly model-dependent, making the comparison between studies difficult. For instance, 3-month-old TG2576 mice showed early hypersynchronisation in the hippocampus, followed by a hypersynchronisation in the DMN, the thalamus, the cingulate and piriform cortices 2 months later. These alterations of brain networks were demonstrated to be associated with an increased ratio of excitatory/inhibitory synapses [
45]. However, in PDAPP mice, an early hypersynchronisation was only found in the frontal cortex at 3 m of age. The variability observed across the mouse models of amyloidosis could originate from model specific characteristics. Accordingly, in APP-overexpressing transgenic mouse models of amyloidosis, the mutant APP is either the human transgene or a chimeric mo/huAPP transgene bearing humanized Aβ domain with single or multiple fAD-associated mutations [
17]. Independent of the type of mutations incorporated in the transgene, the levels and start of APP transgene overexpression might also interfere with the observed phenotype [
28]. Although the role of APP is still not fully understood, APP was shown to be involved in synaptic functions and plasticity [
36,
58] as well as in GABAergic neurotransmission, at pre- and postsynaptic level under normal physiological conditions [
53]. On the contrary, an overexpression of APP could lead to an opposite effect, being toxic for the neurons. Furthermore, the promotor used to construct the transgene may also have an impact, inducing strong and persistent transgene expression not per se restricted to neurons, or to the brain, with some active in utero. Some of these effects have been reported in the literature, e.g. a persistent locomotor hyperactivity in adult mice [
43], an overall reduction in the number of adult-generated hippocampal neurons [
54], cognitive deficits and pathological features unrelated to Aβ levels [
49] or an abnormal electroencephalogram (EEG) activity [
9]. However, the early intra- and internetwork FC hypersynchronisation seems to be due to increased sAβ levels rather than the effect of APP overexpression, as demonstrated in the knock-in APP
NL-F/NL-F mouse model of amyloidosis expressing APP at physiological level, which exclude the interference of APP artefacts [
46]. Interestingly, a similar phenomenon was observed in humans. Asymptomatic young adults with fAD showed greater activation in the right anterior hippocampus compared to matched controls [
41]. Cognitively-intact children at genetic risk for fAD showed an early hypersynchronisation of the DMN, which coincided with significant higher plasma levels of Aβ
1–42 [
42]. Furthermore, healthy young adult APOE-ε4 carriers (known as greatest genetic risk factor for sAD) displayed increased activation in the DMN [
19,
48] and in the hippocampus [
50]. This hyperactivation seems to be persistent in the hippocampus of older adult APOE-ε4 carriers [
8]. Interestingly, at 16w post DOX treatment, the TG mice showed similar FC values as compared to the controls. This decreased FC comparable to controls’ FC might be explained by the decreased of sAβ levels, as the Aβ plaques started to deposit around this time point as confirmed by Thioflavin-S staining of brain slices (data not shown), suggesting that the plaques might be inert, protecting the brain from sAβ toxicity, as a dynamic equilibrium might exist between toxic oligomers and inert plaques [
5].
Nevertheless, the continuous production and accumulation of sAβ revealed to exacerbate its toxicity after the wide-spread deposition of amyloid plaques in cortical areas and hippocampus. Indeed, at 28w post DOX treatment, an overall hyposynchronisation within and between brain networks were observed in the TG mice, and more specifically in the DMN. These findings are consistent with previously reported studies using developmental-onset amyloidosis models in which significantly weaker intra- and internetwork FC was observed at advanced stages of Aβ pathology [
44‐
46]. In effect, 18-month-old APP/PS1 mice (plaques deposition starting at 2 m of age) showed weaker FC across the entire brain, as well as lower interhemispheric FC in the somatosensory cortex and the hippocampus at advanced stages of Aβ pathology [
44]. In the knock-in APP
NL-F/NL-F mice (plaques deposition starting at 6 m of age), a telencephalic hyposynchronized FC was found at 7 m of age, as well as hyposynchronized FC between hippocampus and prefrontal cortex, striatum and hippocampus, and striatum and cingulate cortex [
46]. However, the TG2576 and PDAPP mice demonstrated hyposynchronisation of brain networks before Aβ plaque deposition, which worsened with age [
45]. These observations are in agreement with human data. Indeed, patients with mild cognitive impairment and AD patients demonstrated to have a consistent decreased FC in the DMN compared to healthy controls [
23,
47,
56]. Interestingly, the regions prone to sAβ-induced hypersynchronisation in absence of plaques tend to deteriorate into subsequent hyposynchronisation after Aβ plaque burden. This decrease connectivity in the neuronal networks seems to be caused by a progressive sAβ-induced synaptic damage.
Since sAβ is continuously produced and amyloid plaques might also serve as a reservoir, the subsequent gradual and persistent synaptic loss reflected by first a hypersynchronisation followed by a hyposynchronisation of neuronal networks might come from a failure of the homeostatic mechanisms [
52]. Indeed, the accumulation of sAβ oligomers in the extracellular space could reflect a deficiency in metabolites clearance or in vascular hemodynamic responses, as an increased cerebral blood flow seems to compensate the increased amyloid burden in cognitively normal Aβ + elderly [
18]. Thus, the synapse silencing might be a compensatory mechanism to counteract sAβ-induced hyperactivity of the neurons, leading however to memory deficits at advanced stage of AD. As a change in synaptic transmission provide a physiological substrate for cognition, it is not surprising that the DMN is vulnerable and affected by sAβ during AD progression [
11]. The DMN network includes regions which are closely involved in cognitive functions, such as the hippocampus in learning and memory processing [
55]. However, its capacity to neurogenesis might be limiting for the homeostatic regulatory system in stabilizing this hub of plasticity [
52]. Indeed, adult hippocampal neurogenesis is sharply reduced in AD patients [
34]. Thus, hippocampal networks are more prone to become disturbed early in AD than the primary sensory cortices, which remain fully functional until late stages.
While further experiments are needed to shed light on the mechanisms underlying the early hypersynchronisation of neuronal networks in AD pathology, the inducible Tet-Off APP mouse model of amyloidosis could be an interesting and useful tool when used as mature-onset. The identification of the local concentration of most toxic sAβ species would provide crucial information on the early events triggering the pathology. Overall, the evaluation of the efficacy of the clearance mechanism and vascular responses will give insight on the possible causes for sAβ accumulation.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.