Neuronal Hyperexcitability in APP SWE /PS1dE9 Mouse Models of Alzheimer’s Disease

INTRODUCTION

Alzheimer’s disease (AD), the most common cause for dementia in old age, is characterized as a neurodegenerative process that usually progresses for years and manifests itself in cognitive dysfunction and behavioral changes. According to the amyloid cascade hypothesis, cerebral amyloidosis, which is mainly caused by an imbalance of production and clearance of amyloid peptides, especially amyloid-β (Aβ), is still considered a main driver of AD pathogenesis [1]. In this process, ubiquitously expressed amyloid-β protein precursor (AβPP), which is particularly elevated in neurons and synapses, is a key factor [2]. Physiological cleavage of AβPP is mainly mediated through α-secretase which produces soluble amyloid peptides, whereas AβPP processing in AD is mainly mediated by sequential cleavage through β-secretase and γ-secretase, containing presenilin (PS) 1 or 2 as core proteins, contributing to the production of insoluble amyloid peptides. Impaired AβPP proteolysis, caused by mutations in APP or PS genes, or impaired Aβ clearance drive the build-up of insoluble forms and oligomerization of amyloids. A primary dysbalance of Aβ production and clearance leads to downstream synaptic and neuronal network dysfunction, eventually manifesting as electrophysiological and clinical signs of cortical hyperexcitability [3], and ensuing neuronal loss and cognitive and behavioral impairments.

Alois Alzheimer already observed seizures when first describing the symptoms of a patient with the disease that was later named after him [4], but the possible key role of abnormal network activity for progression of cognitive and behavioral symptoms in AD was only noticed much later [5, 6]. In this concept, subclinical epileptiform activity and seizures represent consequences of upstream amyloid pathology [7–9], which could even be observed in early disease stages [10, 11]. At the same time, cortical hyperexcitability serves as a driver of local amyloid production, providing one potential mechanism feeding the dynamic of the amyloid cascade. This interplay is illustrated schematically in Fig. 1 and will be further developed throughout the review. Consistent with such a role, clinical studies found that hyperexcitability was associated with faster cognitive decline in AD cases [10, 12]. Nevertheless, literature shows a high variability of incidence and prevalence of epileptic activities in AD [13, 14]. Subclinical epileptiform activities are easily missed and amnestic or confusional episodes related to epileptic activities are easily mistaken for symptoms found in AD [9, 15]. Additionally, investigations of aberrant neuronal activities in humans mostly rely on analysis of EEG recordings of the whole brain and lack information of neuronal activities in smaller circuits or on single cell level. This points to the need of investigating the role of hyperexcitability in AD pathogenesis in suitable preclinical models.

OPEN IN VIEWER

Fig. 1

One basic concept for understanding the pathogenesis of Alzheimer’s disease (AD) is the strong association between neuronal hyperexcitability and cerebral amyloidopathy, which are in constant, mutually reinforcing interaction. Investigation of modulating factors in this process may help to further elucidate pathogenic mechanisms in AD.

In the last decades, transgenic mouse models of cerebral amyloidosis served as one key approach to investigate disease mechanisms of AD, including cerebral amyloidopathy, characterized by oligomerization of soluble amyloids and generation of amyloid plaques (reviewed in [16]) and cerebral tauopathy, characterized by hyperphosphorylation of tau and generation of neurofibrillary tangles (reviewed in [17]). Furthermore, different downstream features of primary cerebral amyloidosis have been successfully replicated in transgenic animal models, including subclinical epileptiform activities emerging from neuronal hyperexcitability [18]. Furthermore, using transgenic models to investigate newly formed pathogenetic hypotheses can provide biomarkers for predicting cognitive change, validating evidence of interventional effects on cerebral hyperexcitability, as well as for characterization of targets for directed interventions. One very promising domain of such translational biomarkers are electrophysiological recordings in animal models and humans. Thus, future research with AD models can help to find treatment options to arrest or delay disease progression and to identify translational markers to measure treatment effects on disease pathogenesis.

In the following section, we will characterize the broadly used APPswe/PS1dE9 models of cerebral amyloidosis, present current state of the art methods for determining changes in neuronal excitability and activity, and report the resulting findings. A specific focus will be on the interaction of cerebral amyloidosis with excitability, changes in sleep-wake-rhythm and malfunction of glymphatic clearance. Due to the focus of this work, we do not claim to give full account of the translation of the preclinical findings to AD pathogenesis in human studies. Still, we point out at several instances how studies in humans may benefit from the preclinical results.

APP_SWE/PS1dE9 MICE AS MODELS FOR AD

Following the amyloid cascade hypothesis [19, 20] and observations that every known inherited form of AD is related to impaired amyloid production or clearance, the large majority of AD models focusses on Aβ pathology based on known mutations in the APP and/or PS genes related to familiar forms of AD with early onset [21–23].

Researchers established ample unique knock-ins of mutated human or chimeric forms of APP or PS genes and their combinations in distinct mouse lines (https://www.alzforum.org/research-models). However, combinations of different human APP mutations, especially the Swedish mutation, paired with different human PS1 mutations excel in number and popularity. Examples include PSAPP mice [24], 5XFAD mice [25], and ARTE10 mice [26]. The mouse lines mostly differ in the specific mutations, but also in the used promoters and genetic backgrounds. In current research, double transgenic APP/PS1 mice are well-known and broadly used, as the synergistic effect of mutated APP and PS1 facilitates an early onset of detectable Aβ pathology and a more rapid course of disease compared to monogenic lines [27]. So far, different AD mouse models only reflect partial aspects of the highly complex disease. Therefore, choosing the best matching model for the designated research hypothesis is crucial [28, 29].

Commonly used APP/PS1 models are APPswe/PS1dE9 strains [30] that harbor both the Swedish mutation KM594/5NL in a chimeric mouse/human APP gene and a deletion of exon 9 in the human PS1 gene, that both initiate the accumulation and aggregation of Aβ [31]. This model was originally created and maintained on a hybrid C57BL/6×C3H background [31], but subsequent studies used mice backcrossed to a C57BL/6 line [18, 32–35]. In both backgrounds, the transgenes are expressed under the mouse prion protein promoter which drives expression in the central nervous system (CNS) [31]. As characteristic hallmarks of AD, this model shows cerebral amyloidopathy [36] and behavioral as well as memory deficits [37, 38]. Of specific interest for the current review, this model also demonstrates changes in neuronal activity leading to subclinical epileptiform activity as well as to seizures [33, 39], which in contrast to many other AD models even generalize and are associated with observable epileptic motor signs [11]. The combination of well-defined primary cerebral amyloidosis and downstream changes of neuronal function and excitability allows studying potential mechanisms of propagation of amyloid effects to neuronal network dysfunctions as well as possible feedback effects from increased synaptic activity on amyloid production. As an important modulator of this interaction, mechanisms of brain clearance related to circadian glymphatic system integrity are accessible to studies in APPswe/PS1dE9 mice [40–42].

We are well aware that many other transgenic models provide valuable data on the association between cerebral amyloidosis and hyperexcitability as well, like the hAPPJ20 model containing the Swedish and Indiana mutation of APP [6, 43] or the APP23xPS45 model containing the Swedish APP mutation and the G384A mutation in PS1 [44, 45]. Describing findings across all these models would go beyond the scope of this review. Therefore, we focus on APPswe/PS1dE9 strains in the following sections, as the described pathological features characterizing APPswe/PS1dE9 mice underline the superior applicability of this model in research on the complex relationship between cerebral amyloidosis and neuronal hyperexcitability.

Despite promising features of currently available transgenic models, including APPswe/PS1dE9 mice, several factors limit the transfer of findings to AD in humans. As one key point, cerebral amyloidosis in these transgenic models resemble familiar forms rather than sporadic forms of AD. At the same time, familiar forms represent less than 1%of AD cases [46]. In autosomal dominant forms of AD due to APP or PS mutations, overproduction of insoluble Aβ species is considered the primary pathogenic event, whereas in sporadic AD disrupted Aβ removal through altered brain clearance mechanisms may be the key driver of pathology [1]. Furthermore, the incapability of transgenic APP/PS1 models to reproduce characteristic AD-like neurofibrillary tangles and neuronal cell loss is a well-known problem [47, 48]. Therefore, mouse models of AD are often called incomplete and improved models reproducing a complete pathological cascade of AD would be valuable [27, 47], but are clearly out of reach. Despite this incompleteness, these models allow testing specific pathogenic pathways, like neuronal hyperexcitability and examining the interrelationships between different pathways and lesions [47], if the experiments are embedded in a well-defined conceptual framework.

ELECTROPHYSIOLOGICAL MEASUREMENTS TO INVESTIGATE NEURONAL NETWORK DYSFUNCTION

Currently there is a wide spectrum of available methods for observation of neuronal hyperexcitability, neuronal hyperactivity, and disruptions in neuronal networks in mouse models of AD. This article cannot give a full description of these methods, but only highlights some of the most commonly used experimental procedures, to enhance the comprehensibility of the presented results in the following paragraph.

On the single cell level, one can monitor synaptic function and single neuron properties. One approach is to perform voltage-clamp analyses [49] to examine voltage-dependent channel currents of single neurons while holding their membrane voltage at a set level. Using this method, increased sodium channel current densities were assessed in APP/PS1 mice [35]. A possible method to observe membrane resistance and the cellular response to induced currents are current-clamp analyses. For instance, this technique may evidence intrinsically hyperexcitable burst firing patterns depending on Ca²⁺ and K⁺ conductances [50]. However, in APPswe/PS1dE9 mice, this has not been addressed so far. When focusing on neuronal communication on a between-cells level, patch-clamp analyses as well as sharp electrodes or field potential recordings are performed to observe input-output-curves and synaptic plasticity. Numerous studies found impaired long-term potentiation (LTP) in the APP/PS1 hippocampus in vitro [51, 52] and in vivo [53]. By expanding from the cellular to the network level, large neuronal circuits are investigated by electroencephalographic (EEG) recordings, combined with video (video-EEG) and subsequent power-analysis or assessment of the spike load to determine effects of AD pathology on large-scale hyperexcitability. This is also the only of the presented methods that is commonly used to study epileptiform activities in humans. Due to the cognitive deficits in AD patients, it is often not possible to obtain a valid clinical history. This renders mobile video recordings and postictal EEGs elementary tools for differential diagnosis to distinguish between symptoms arising from AD or epileptic activities [13]. In principle, insights gained with EEG measurements can be compared between mice and humans and add an interesting translational potential to preclinical research on neuronal hyperexcitability in AD models.

In addition to electrophysiological methods, other methods allow to determine properties of neuronal activity in AD mouse models. Since Ca²⁺ transients directly reflect action potential firing in neurons, fluorescence labelling of Ca²⁺-ions enables assessment of Ca²⁺-signaling through in vivo or in vitro assessment with fluorescence-microscopy in order to simultaneously study a large number of neurons [44, 55]. These investigations allow assessment of synchronization on a cellular level.

Finally, the field of optogenetics is a new promising method for investigating neuronal hyperexcitability. Optogenetics uses genetic changes that are activated with light of defined wavelengths to control cellular activities [56]. When studying neuronal hyperexcitability, optogenetics enable interventions and methods to observe how individual neurons behave when excited and how this affects the activity in larger neuronal networks [57, 58]. Even if optogenetics is not yet widely used, it is a highly promising method for future studies of hyperexcitability in the course of AD. Therefore, it should be considered as valuable tool for further investigations in APPswe/PS1dE9 mice and other models.

Aβ-DEPENDENT CHANGES IN EXCITABILITY OCCUR AT DIFFERENT ORGANIZATIONAL LEVELS IN APP_SWE/PS1dE9 MOUSE BRAINS

APPswe/PS1dE9 mice develop AD characteristic Aβ pathology reliably [31, 36] and replicate downstream Aβ-dependent lesions and disruptions in neuronal excitability and neuronal activity. Such changes manifest themselves in patients as subclinical epileptiform activity in EEG recordings and clinical evident seizures [7, 11]. Reported hyperexcitability derived changes in patients therefore are mostly focused on whole brain network activity. Unfortunately, underlying mechanisms of neuronal hyperexcitability in AD pathogenesis and disease progression remain elusive. Researchers often used APPswe/PS1dE9 models in preclinical investigation of changes in neuronal network function on different organization levels in the CNS. For a structured insight, we review the reported changes according to the different organizational levels of the brain.

The smallest functional subunits involved in cerebral signal transmission are synapses. It is already known that synaptic loss is highly associated with cognitive impairment [59–61]. However, it has not yet been well characterized which structural and functional brain changes during the course of AD are leading causes of synaptic loss and which changes only emerge after synaptic loss already has taken place. Synapses for neuronal signal transmission are functionally dependent on receptor function mainly relying on two components: ion channels and transmitters. While APP deletion was associated with impaired synaptic plasticity [62], transgenic APPswe/PS1dE9 mice showed changes in ion currents involved in signal transmission and action potential generation and in their associated channels, with a noteworthy strong increase in the neuronal expression of voltage-gated sodium channels, like Na_v1.6 [35]. In addition, functional studies found pre- and postsynaptic cholinergic deficits and a generalized decreased neuronal responsiveness to acetylcholine (ACh) [34, 63], a neurotransmitter that is known as key-mediator in processes of memory formation. Its malfunction is closely linked to AD [64]. Another neurotransmitter closely linked with neuronal excitability is gamma-aminobutyric acid (GABA). Activation of its receptors leads to chloride-ion currents, which usually have a hyperpolarizing effect on the cell membrane and thus generate inhibitory signals within the circuitry regulation of neuronal activity. Derailment of this circulatory regulation through reduced levels of GABA-transmitter and -receptors [65] may lead in general to a lack of inhibitory GABAergic synapses and in larger scale also GABAergic neurons. This can drive neuronal hyperexcitability, as shown in APPswe/PS1dE9 mice with experimental verified loss of GABAergic synapses and neurons consistent with the notion that loss of inhibitory GABA function facilitates neuronal hyperexcitability [66, 67]. Consistently, optogenetic restoration of GABAergic function led to an improvement of hyperexcitability shown on higher organization levels as normalizations in EEG band power in APPswe/PS1dE9 mice, furthermore even leading to ameliorated memory function [68]. The disease modifying effect of GABAergic neuron function restoration through progenitor cell transplantation was also reported in APPswe/PS1dE9 mice [67].

A mechanism observed on synaptic level downstream to neuronal hyperexcitability is synaptic sprouting, a process first described in the pathogenesis of temporal lobe epilepsy [69, 70]. It is reported as a process of synaptic reorganization resulting from changed neuronal activity or from neuronal loss, where an excessive activity leads to a pathological rearrangement of neuronal circuitry that manifests as increased excitability (reviewed in [71]). Similar sprouting was also found in APPswe/PS1dE9 mice with hyperexcitability-derived seizures [33] underlining convergent mechanisms of hyperexcitability from different upstream pathologies, including AD characteristic cerebral amyloidosis.

Changes in ion channels and transmitter sensitivity and release in the APPswe/PS1dE9 AD models lead to a reduced action potential threshold on synaptic level [33, 35] and thus affect individual neurons as next higher organizational structure. Changes in synaptic neurotransmitter activity and ion currents lead to membrane depolarization and spontaneous membrane potential fluctuations that both enhance neuronal excitability and in turn favor higher firing rates up to burst firing activity of neurons [33, 39]. The impact of neuronal hyperexcitability on the whole network differs regarding which specialized neurons it affects, for example regulatory interneurons play a central role. It was shown that Aβ is able to induce hyperexcitability in regulatory interneurons in APPswe/PS1dE9 mice, which leads to inhibited neuronal transmission, manifested as deficits in spatial learning and memory typical for AD [72]. Those results underline that hyperexcitability-derived disturbance in well-orchestrated interneuronal connections lead to malfunctions in neuronal circuits and networks, as highest organization levels in CNS.

On this high organizational level, transition failures become apparent as brain network fragmentations [39] and changes on synaptic and single neuron levels leading to a generalized excitation-inhibition-imbalance [32]. Among all of the presented methods EEG is the preferably used method to examine neuronal activity of larger circuits in the brain in patients as well as AD models. Neuronal network dysfunctions lead to measurable changes in EEG recordings. Surprisingly, most EEG-recording results in APPswe/PS1dE9 mice described increases in alpha and beta power as well as decreased delta and theta power [32, 74], which are contrary to abundant observations in AD patients on increasing slow-wave activity [10, 76]. Only one research group recently described decreased alpha and increased delta power in hippocampal EEG-recordings of APPswe/PS1dE9 mice matching similar findings in human AD [77]. Unfortunately, relative EEG delta and theta power mainly depending on the mouse strain and were for example observed to be less expressed in C57BL/6 mice compared to BALB/c mice [78]. Along the same line, focal epileptiform activity following administration of penicillin was less pronounced in C57BL/6 mice than in BALB/c [78]. This makes it obvious that changes on the network level are not easily transferable between transgenic models and human studies. Transfer not only depends on the actual transgene but also on the background strain.

EEG analyses also allow observation of murine gamma brain activity generated in the hippocampal-entorhinal network [79]. Interestingly, APPswe/PS1dE9 mice exhibited an increase in hippocampal gamma power compared to wildtype [80]. At first glance, this opposes observations of reduced gamma activity in older people [81]. In AD patients, however, EEG results regarding gamma activity are more controversial [82] with increased gamma power and connectivity in some clinical studies [83, 84].

When looking on methodical details of studying neuronal hyperexcitability on different organizational levels, the cited experiments were performed on mice of different ages (4 months-16 months), at different activity levels (awake, non-REM sleep, during Y-maze task) and with focus on different brain regions (cortex, hippocampus, thalamus) [32, 80]. This leads to difficulties when comparing the results with each other and with insights from clinical studies and point to the need for a higher standardization to increase comparability.

Additionally, researchers identified a higher rate of epileptic and epileptiform activity in EEG-recordings in an impressive number of studies in APPswe/PS1dE9 mice associated with spiking-activity, spike-wave discharges or a twisting appearance of EEG records [18, 73]. The high correlation of cerebral amyloidopathy and neuronal hyperexcitability in APPswe/Ps1dE9 mice, which manifests for example as epileptiform activities, suggested that antiepileptic medication may be promising to ameliorate AD-like symptoms in preclinical studies [18]. Indeed, in APPswe/PS1dE9 mice antiepileptic drugs like carbamazepine, phenytoin, and valproic acid, all acting through inhibition of Na⁺-currents, reduced epileptiform discharges [18]. Another antiepileptic drug used in in vivo studies is levetiracetam, which was shown to reduce spontaneous seizures and to have beneficial effects on cognition in APPswe/PS1dE9 mice [85]. Subclinical epileptiform activity as pathologic changes could also be shown in AD patients [7, 10]. Consequently, several clinical trials with antiepileptic medication, e.g., levetiracetam, emerged [86] (ClinicalTrials.gov Identifier: NCT01044758); [87] (ClinicalTrials.gov Identifier: NCT01554683). Published results showed improvements in hippocampal hyperexcitability and cognitive function in patients with amnestic mild cognitive impairment, an early stage of AD [86]. In patients with mild AD antiepileptic medication with levetiracetam showed a beneficial effect on neuronal hyperexcitability, such as normalizations of EEG band power spectra, but no significant effect on cognitive function [87]. Currently, numbers of cases to evaluate the disease-modifying effect of antiepileptic medication like levetiracetam, lamotrigine and phenobarbital in phase II clinical trials, like [88] are still low, rendering the results on potential disease modifying effects of antiepileptic medication based on the hyperexcitability mechanism of AD propagation inconclusive. Evidence on efficacy from phase III trials is still widely lacking.

Moreover, as shown in preclinical studies, neuronal hyperexcitability in larger neuronal networks is closely linked with processes like changes in circadian activities and sleep-wake-rhythm. Circadian changes in body temperature and phase delay of daytime activity were shown in APPswe/PS1dE9 models [89, 90]. These findings are reminiscent of similar observations in AD patients, such as agitation and arousals in afternoon or evening, so called “sundowning” [91], disturbances in clock-gene regulation [92], shift in sleep-wake cycles [93], and dysregulation of body temperature [94]. Conversions of sleep-wake-cycles shown in animal models and clinical studies in AD patients are likely not just an epiphenomenon of the disease but also drivers of the pathological cascade [95–97], among other mechanisms through impaired cerebral clearance of pathological proteins [98, 99].

However, it is important to note that the reported pathological changes (as summarized in Table 1) differ in severity depending on the age and genetic background of the models. Particularly, changes in genetic background tend to have an impact on disease phenotype, as properties like response to neurodegeneration and excitotoxicity as well as cognitive abilities in general are strain-dependent [27]. Therefore, results suggest that hyperexcitability and changes in neuronal activity are not only a function of age or the transgene but also depend on the background genotype [39]. Again, this supports the notion that researchers need to be cautious when comparing results across different AD models and when transferring knowledge from preclinical studies to AD patients.

HYPOTHESES FOR UNDERLYING PATHOGENIC MECHANISMS OF HYPEREXCITABILITY AND THEIR REGULATORY ROLE IN AD

As it became apparent in the previous section, neuronal hyperexcitability and cerebral amyloidosis are intertwined in complex and multilayered ways. Based on data obtained from AD models, including the APPswe/PS1dE9 mice, complemented with data from clinical studies, one can construct a model of derailed physiological functions within a multifactorial and multidirectional vicious cycle that contributes to the manifestation of AD (Fig. 2). As the underlying causes of AD remain elusive, results of preclinical studies can be interpreted in the context of different, partly competing hypotheses. This allows forming a more comprehensive overall picture for understanding disease mechanisms as key basis to refine diagnostics and identify new therapeutic targets.

OPEN IN VIEWER

Fig. 2

Cerebral amyloidopathy is an essential AD characteristic (dark highlighted) and seems to be the pathogenic basis for further Aβ-mediated processes. The diagram illustrates interconnections and feedback mechanisms linking amyloidopathy with neuronal hyperexcitability (directly leading to increased neuronal hyperactivity) and disruptions in circadian function (directly leading to disrupted sleep-wake-rhythmicity) in AD pathogenesis. Feedback mechanisms of downstream pathways promote increasing imbalance of Aβ production and clearance amplifying amyloidopathy and downstream mechanisms promoting AD.

According to amyloid cascade hypothesis [19, 20] cerebral amyloidopathy is the most important trigger for other dysfunctions in AD pathogenesis. The cause for amyloidopathy is an imbalance of Aβ production and clearance, resulting in Aβ deposits called Aβ plaques. Downstream of amyloid pathology, hyperexcitability is leading to increased neuronal activity [44, 45]. Elevated Aβ levels seem to be causally linked to epileptiform activities in this context [43]. At the same time, in a feedback loop increased neuronal activity leads to an increased formation of Aβ by increased pathogenic cleavage of AβPP [58, 100].

As mechanistic basis connecting AD and neuronal hyperexcitability, the presented research data point to the importance of neurotransmitters in synaptic mediation. Observation of pathomechanics in AD from a molecular point of view shows that neuronal loss is mediated through excitotoxic effects of Aβ, exerted as ion-dyshomeostasis [101] and aberrant neurotransmitter actions [102, 103]. AD especially affects cholinergic interneurons with an enormous impact on their modulatory functions in neuronal circuitry (reviewed in [104]). This is in line with the ACh-hypothesis, describing symptoms of AD originating from decreased choline acetyltransferase activity and loss of cholinergic neurons, e.g., in the nucleus basalis of Meynert [105, 106], leading to neuronal malfunctions arising from disrupted neurotransmitter signaling. Interestingly, choline supplementation in APPswe/PS1dE9 mice was shown to restore impaired neuronal function and improve spatial learning and memory [107]. As ACh exert its effects through nicotinic ACh-receptors located in pre- and postsynaptic membranes, modulating the release of itself and numerous other neurotransmitters like glutamate or GABA, ACh directly influences synaptic activity and neurotransmission [104]. Together with neurotransmitter signaling, synaptic dysfunction is recognized as anatomical correlate of cognitive impairment in AD [76, 108] that promotes neuronal death and macroscopic morphological changes in AD brains leading to further functional impairments.

Directly related to these observations, the ‘Synaptic Failure’-hypothesis describes malfunctions of the synapses to be initially, directly or indirectly, triggered by Aβ [44, 109]. Emanating from impaired synaptic signaling and neuronal transmission, there is a growing number of scientific works showing disruption of the excitation-inhibition balance leading to disturbances of activity in larger neuronal circuits and networks [33, 110]. It is apparent that disrupted functionality at lower organization levels of the brain determine malfunctions in higher levels and vice versa. From a point of view more oriented on neuronal functionality, those results are interpreted as reflections of hyperexcitability and a possible explanation for the causal relationship between AD and an exacerbated vulnerability to seizures up to increased epilepsy prevalence in AD models [33, 34] and patients [7, 11]. On the other hand, observed neuronal hyperexcitability and hyperactivity initiate compensatory inhibitory processes; these processes normalize aberrant activities in hyperactive networks; however, at the same time they may trigger hypoactivity in other neuronal circuits leading to learning and memory deficits [5].

Aβ-derived hyperexcitability highlights again the central role of Aβ in AD and corroborates the amyloid cascade hypothesis [19, 20]. Aβ peptides formed in the process of AβPP cleavage, unfold their toxic potential on downstream mechanisms, finally leading to neuronal death [16]. Aβ molecules cause synaptic dysfunction in a direct way or trigger amyloid-dependent release of proinflammatory mediators, like reactive oxygen species or cytokines [44], promoting neuroinflammation and oxidative stress, which further damage synapses and neurons exaggerating neuronal loss. Experiments in APPswe/PS1dE9 mice point to oxidative stress mediating effects and cytotoxic effects of Aβ-molecules that impair neurogenesis from progenitor cells and further exaggerating neuronal loss [111].

Possible mechanisms for neuronal hyperexcitability to drive disease progression are feedback effects, where amyloid-induced neuronal hyperactivity increases amyloid formation [58, 100]. Additionally, abnormal neuronal network activity interferes with amyloid clearance mechanisms of the brain via disturbances of the sleep-wake-rhythm and the closely related functionality of the so called glymphatic system [112, 113]. The term glymphatic is a neologism formed of the terms glia-mediated and lymphatic, which describe main aspects of CNS fluid interchange. In contrast to the rest of the body, the brain lacks a supply of lymphatic vasculature. However, the drainage of extracellular fluids in the brain is essential for fluid volume balance and interstitial waste removal. To substitute lymphatic system function in the brain, glia-mediated perivascular networks have formed and work as the glymphatic system (reviewed in [114]). Understanding the reciprocal interaction between cerebral amyloidosis and neuronal excitation and its influence on sleep-wake-rhythm and glymphatic clearance is necessary for future AD investigations. Along this line, increased cerebral concentrations of pathologic molecules through changes in glymphatic mechanisms of the brain probably play another central role [98, 115], as physiological clearance of detrimental mediators like Aβ, reactive oxygen species, or cytokines is disrupted. As glymphatic function is dependent on circadian function [113], a mechanistic linkage between hyperexcitability, sleep, and AD becomes overt. Sleep disorders and sleep disturbances are well known risk factors for AD [95, 96]. Interestingly, in humans with epilepsy a strong correlation between aberrant neuronal activity, sleep disorders [116], mood and behavioral changes [117], and cognitive decline [118] similar to observations in AD was also shown. These are clues for hyperexcitability as converging pathomechanism in epilepsy and AD showing its interconnection with dysfunctions in neuronal regulation of further physiological processes in the form of different bi-directional relations or even feedback loops [119, 120]. As one example for existing feedback mechanisms, decreased Aβ concentration was shown to be associated with decreased wakefulness in APPswe/Ps1dE9 mice [121], mechanistically linking AD and sleep disturbances. Reduced sleep quality and quantity lead to an increase in hyperexcitability-characteristic signaling pathways [122] and facilitate Aβ deposition through disruption of circadian rhythm in brain clearance [112]. Whereas impaired glymphatic clearance is one major reason for Aβ plaque formation, Aβ plaques are equally able to impair glymphatic clearance themselves, especially when they deposit in cerebral arteries or small capillaries [123]. Interestingly, glymphatic influx was correlated to high EEG delta power [124] in APPswe/PS1dE9 mice. Impaired Aβ clearance seemed to be directly linked to further pathogenic processes leading to behavioral and cognitive changes [42] despite decreased delta power ([32, 74], but see [77]). On the other hand, restoration of CNS Aβ clearance was shown to be correlated with restored cognitive function [40] in mice. This points to the possibility to use measurements of Aβ clearance via the glymphatic system as a possible driver of cognitive decline.

Based on inherited genetic modifications of APPswe/PS1dE9 mice, amyloidopathy through an initial increase of Aβ production seems to trigger further downstream mechanisms. Among these, neuronal hyperexcitability is paramount, which itself causes further pathomechanisms finally leading to cognitive dysfunction but at the same time exacerbating amyloid-pathology through feedback mechanisms. This is also likely to be a possible mechanistic basis in familiar forms of AD. However, in the more prevalent sporadic AD the question which of amyloidopathy and hyperexcitability is cause and effect is a chicken or egg causality dilemma. Key findings, especially from AD mouse models help to unravel the complex interconnections between amyloidopathy, hyperexcitability, and circadian function. Each of those features is able to set a process in motion exacerbating itself and other downstream mechanisms in AD progression forming a vicious circle. This underlines the thesis that AD progression is not just a cascade but rather a much more complex process [125].

FUTURE BENEFITS FROM THE KNOWLEDGE ABOUT HYPEREXCITABILITY IN THE APP_SWE/PS1dE9 MODEL

It remains unclear which factors in the presented circulatory relationship between amyloidopathy and hyperexcitability are consequences of the other pathologic change. Therefore, this topic is still a promising field of future research with the aim to identify key mediators, which determine AD progression and may serve as biomarkers or new therapeutic targets.

Neuronal hyperexcitability has already been studied on different organizational levels of the brain in the broadly used APPswe/PS1dE9 mice. So far researchers have focused on hyperexcitability alone, but not in concert with the related phenomena of sleep-wake-rhythms or circadian brain clearance. A possible approach to such combined investigations is the use of non-invasive state-of-the-art methods in APPswe/PS1dE9 mice. Transferability of those insights remains a major difficulty of preclinical research [126, 127]. To increase transfer of findings, methods like multimodal in vivo imaging and EEG recordings, commonly used in humans, should be preferred [128]. To further elucidate the role of hyperexcitability in the complex vicious cycle contributing to AD, imaging should be combined with established EEG measurement methods in mice [18, 73], originating from diagnostic techniques in humans, which already generated extensive insights on cerebral hyperexcitability in AD patients [129, 130]. As an example, using established methods for in vivo recordings of EEG signal, enabling assessment of hyperexcitability as well as sleep-wake cycles [131], and amyloid load using PET imaging [132], one can determine associations between cerebral amyloid accumulation and hyperexcitability, both in humans and in animal models. Similar effects in studies with such combinational approaches between transgenic models and humans could then mechanistically be elucidated by subsequent studies also using ex vivo methods. Such combined approaches will enhance significance of animal research and allow transferring the resulting conclusions onto patients with a high degree of credibility and reliability.

Parallel to preclinical experiments, more results from clinical studies based on knowledge obtained from animal models will be available in the future. Promising data showed that blocking hyperexcitability by anticonvulsant-treatment suppresses the abnormal neuronal activity and prevents cognitive deficits in APP/PS1 mice [133, 134]. As previously described, such treatment approaches have also been tested in clinical trials ([86] ClinicalTrials.gov Identifier: NCT01044758; [87] ClinicalTrials.gov Identifier: NCT01554683; [88]). Two other phase III trials with tramiprosate showed no efficacy in the overall population but showed efficacy in a secondary analysis of the subsample of Apolipoprotein E 4 (APOE 4) homozygotes [135] (ClinicalTrials.gov Identifier: NCT00088673, NCT0021776), whereby APOE 4 is recognized for a long time as risk gene in late onset familial and sporadic forms of AD [136]. In addition to already completed trials, further clinical studies on anticonvulsant treatment are being prepared (ClinicalTrials.gov Identifier: NCT03489044; NCT03875638; NCT04004702; NCT02002819) offering new possibilities in AD treatment.

One clinical trial aims to elucidate the pathogenic basis by uncovering the correlation between sleep and epileptiform activity in AD patients (ClinicalTrials.gov Identifier: NCT03923569). Another promising approach showed beneficial effects on memory preventing sleep-loss induced hyperexcitability in APPswe/PS1dE9 mice [122]. Based on observations in preclinical studies like this, more than 40 clinical studies are currently ongoing on the relationship between sleep and AD to investigate functional correlations of sleep and pathological changes in the aging brain (e.g., ClinicalTrials.gov Identifier: NCT04044495; NCT04096261) or to establish new interventions for treatment of AD (e.g., ClinicalTrials.gov Identifier: NCT03455569; NCT03075241).

CONCLUSION

AD is an enormous burden on the global healthcare system but also devastating for individual patients and caregivers. A main obstacle for the advancement of treatments is lack of understanding of disease pathogenesis. The currently available models of AD reflect only aspects of the disease but not the whole process, limiting the use of models to elucidate the disease pathogenesis, but allowing studying individual elements of it. This requires the embedding of animal studies into an explicitly formulated hypotheses framework where not the entire model but crucial parts of it can be tested. Besides Aβ plaques, neuronal hyperexcitability, mainly displayed as subclinical epileptiform activity or clinical overt seizures in patients is another hallmark of AD as Aβ pathology and hyperexcitability in AD are entwined in a complex manner. Further investigation of AD pathogenesis in animal models displaying both Aβ plaques and hyperexcitability is therefore of high value. Suitable, often used and well-described models are APPswe/PS1dE9 mice. In this review we presented evidence for hyperexcitability on different neuronal organization levels, from synapses to larger neuronal circuits, that emerged from these models. Researchers found signs for abnormal synaptic function and synaptic reorganization, changes in membrane threshold and spontaneous neuronal firing activities leading to excitation-inhibition imbalances in larger neuronal circuits and networks. By interpreting the presented data in the context of widely accepted hypotheses, we pointed to potential feedback mechanisms and interconnections between AD, hyperexcitability, circadian rhythmicity, and amyloid clearance. These results suggest that AD and subclinical or clinically manifest epileptiform activity are often occurring as comorbidities due to converging mechanisms finally leading to cognitive and behavioral changes. The findings presented here also provide clues and goals for future research to uncover the pathogenesis of AD and to identify new therapeutic targets.