Introduction
Alzheimer’s disease (AD) is characterized by cognitive decline, and presents histopathologically with amyloid-β (Aβ) deposition in extracellular plaques and intracellular neurofibrillary tangles made up of hyperphosphorylated tau [
1],[
2]. Besides these pathological hallmarks, it is becoming increasingly clear that AD is associated with alterations in neuronal circuit excitability. Patients with Alzheimer’s disease have an increased risk of developing neuronal hypersynchronicity, seizures or other forms of epileptiform activity [
3],[
4]. AD is associated with a 5- to 10-fold increase in seizure incidence, and seizure incidence correlates with lower cognitive performance [
4]. Although seizures were previously thought to be secondary to disease progression, aberrant neuronal activity may directly contribute to cognitive deficits, as neuronal activity appears to regulate regional vulnerability to Aβ [
5]. Modulating epileptiform hippocampal activity modulates cognitive deficits in patients with mild cognitive impairment [
6]. Conversely, forms of epilepsy are accompanied by cognitive impairments [
7],[
8]. Thus, epileptiform activity may contribute to the development of AD-related cognitive deficits. However, the aetiology of epileptiform activity in AD is incompletely understood.
Synchronous network oscillations contribute to normal brain activity, yet, are affected by epileptogenesis [
9],[
10]. In the hippocampus, two main oscillations reflect physiological synchronous activities, both being implicated in behavioural states and memory performance: theta oscillations (4-12 Hz), generated by synchronous phasic firing of pyramidal cells [
11],[
12], and gamma oscillations (25-100 Hz), generated by circuits between GABAergic interneurons and pyramidal cells [
13]. GABAergic interneurons play a key role in synchronization of pyramidal cell firing by providing rhythmic inhibition of pyramidal cells [
14]. In addition, hippocampal interneurons are themselves synchronized by recurrent excitation [
13]. Neuronal networks show further levels of interconnectivity: the phase of theta rhythms modulates the amplitude of gamma oscillations in the hippocampus, and this phase-amplitude cross-frequency coupling (CFC) is critical in hippocampus-dependent memory and cognitive performance [
15]–[
20]. Whether epileptiform activity in AD correlates with aberrations in CFC between hippocampal theta and gamma oscillations is not known. However, aberrant generation of electroencephalographic oscillations has been observed in cognitive impairment and AD [
21], and aberrant hippocampal oscillations have been suggested to occur early in development of AD [
22].
Neuronal expression of human amyloid-β precursor protein (APP) carrying pathogenic mutations has been used to model AD in mice [
23]. APP transgenic mice exhibit extensive aberrant neuronal network and epileptiform activity [
24]–[
32]. Modulating hyperactivity and seizure susceptibility in the hippocampus modulates cognitive deficits in APP transgenic mice [
27],[
31],[
33],[
34]. Hippocampal gamma spectral power, a measure of the contribution of gamma frequencies to the entire signal strength, is markedly diminished in 8 month-old behaving J20 mice expressing mutated hAPP [
35]. Cross-frequency coupling between theta phase and fast gamma amplitude is altered in hippocampal slice cultures from APP transgenic mice [
36]. However, a concise analysis of hypersynchronicity, spectral and CFC aberrations in adult APP mice
in vivo is lacking. Furthermore, the pathways that contribute to network aberrations and hypersynchronicity in APP mice remain incompletely understood.
Oligomeric Aβ may itself affect neuronal circuit excitation [
24]. Aβ reduces excitatory neuronal transmission and plasticity at the synaptic level [
37]–[
41]. Neuronal hyperexcitation and concomitant excitotoxicity in APP transgenic mice require the microtubule associated protein tau [
2],[
33],[
42]. Dysfunction of synaptic NMDA receptors and their downstream signals was shown to underlie loss of inhibitory currents and abnormal hyperexcitation in hippocampal preparations from APP mice [
31]. However, what contributions synaptic NMDA receptors and inhibitory neuron function have in generation and propagation of neuronal network aberrations and hypersynchronicity remains unclear. Furthermore, signalling pathways that may modulate thresholds for aberrant network activity are incompletely understood. Recently, mitogen-activated protein (MAP) kinase p38 has been implicated in the Aβ-induced inhibition of long-term potentiation (LTP) in brain slice cultures [
43]. The role of p38 activity in neuronal network alterations of APP transgenic mice, however, has not been investigated.
In this study, we examined hippocampal hypersynchronicity in adult APP23 transgenic mice using in vivo telemetric electroencephalography (EEG) in free-roaming mice and analyse interictal recording sequences for spectral amplitude distribution and CFC strength before the onset of plaque pathology. We report spontaneous hippocampal hypersynchronicity in APP23 transgenic mice accompanied by marked spectral changes and impaired CFC for theta and gamma oscillations. Furthermore, we addressed thresholds of hypersynchronicity and interictal spectral and CFC distributions upon pharmacological manipulations of voltage-gated sodium channels, which regulate GABAergic inhibition, by riluzole and of NMDA receptors, by the non-competitive inhibitor MK801. Furthermore, we found that MK801 treatment significantly reduces activation of p38 MAP kinase in the hippocampus, and inhibition of p38 alters hippocampal hypersynchronicity thresholds in APP23 mice.
Discussion
In the present study, we report spontaneous hypersynchronicity in APP23 transgenic mice. Furthermore, we show that APP23 mice have marked alterations in theta and gamma oscillations and impaired cross-frequency gamma modulation by the phase of theta oscillations. We found markedly enhanced spike activity in APP23 after riluzole treatment. Surprisingly, we found that hypersynchronicity in APP23 mice could be enhanced by MK801 and the p38 inhibitor SB203580. Spectral analysis showed that enhanced hypersynchronicity in APP23 by riluzole, MK801 and SB203580 is accompanied by alterations in theta and gamma wave power. CFC analysis showed that impaired coupling strength in APP23 mice, while not altered by MK801 or SB203580 treatment, is significantly increased by riluzole (Table
1).
Table 1
Summary of treatment effects on EEG measures in APP23 and non-transgenic controls
Hypersynchronicity | ND |
↑↑
| ND |
↑
| ND |
↑
|
Spike trains per 2 hours (n/120 min) | | 11.6 ± 3.2 | | 6.0 ± 1.2 | | 8.3 ± 2.7 |
Spikes/spike train | | 69.4 ± 3.9 | | 40.9 ± 7.0 | | 12.4 ± 1.1 |
Spike train duration (s) | | 12.38 ± 2.09 | | 6.97 ± 0.74 | | 4.24 ± 0.35 |
Theta power |
↓
|
↑
|
↑
|
↑
|
→
|
→
|
Theta frequency |
↓
|
↓
|
→
|
→
|
↓
|
→
|
Gamma power |
↓
|
↑
|
↓
|
↓
|
↓
|
↓
|
Cross-frequency coupling |
→
|
↑
|
→
|
→
|
→
|
→
|
p38 MAPK activity | nt |
→
| nt |
↓
|
↓
|
↓
|
Spontaneous hypersynchronicity, possible induced by Aβ [
53], has previously been described for transgenic mouse models of AD [
24]–[
32],[
54]. Our study corroborates previous findings by showing spontaneous hypersynchronicity in APP23 mice, which express human APP carrying the Swedish double mutation. APP23 mice present memory deficits as early as 3 months of age [
55], while plaque pathology is not seen before 6 months of age [
44]. Our data show pronounced neuronal network aberrations and EEG abnormalities such as epileptiform activity, spectral power differences and CFC impairments in APP23 mice already at an age when memory deficits manifest, and well before Aβ plaque formation. This is paralleled by altered molecular signalling with hippocampal p38 activation. Neuronal networks may even be compromised before memory deficits develop, since network aberrations and hypersynchronicity have been observed in brain slices of APP-expressing TgCRND8 mice at 1 month of age [
36],[
54].
Spectral analysis showed significantly lower interictal theta power in APP23 EEG recordings, while strength of gamma oscillations was significantly increased (Figure
2). As outlined above, hippocampal oscillations of theta and gamma frequencies contribute to physiological processes required in memory and cognition [
11]-[
13]. Thus, hippocampal network aberrations leading to spectral changes in these frequency bands may underlie onset of cognitive deficits in APP23 mice [
55],[
56]. Alternative APP-expressing mouse models show similar alterations in theta and gamma waveforms [
34],[
36],[
54]. Consistent with these findings in murine models, early human AD pathology and mild cognitive impairment is accompanied by alterations in neuronal network oscillations [
21],[
22],[
24]. Whether altered spectral power of theta oscillations also influence thresholds for epileptiform activity in AD remains unclear. However, data from rodent models of epilepsy suggest that a slow oscillation state affects hypersynchronicity thresholds [
57]. Thus, spectral power changes in theta oscillations may affect hypersynchronicity in APP23 mice.
Hippocampal gamma oscillations are generated by circuits between GABAergic interneurons and pyramidal cells [
13],[
35]. Interestingly, our interictal hippocampal recordings show increased gamma oscillations in APP23 mice compared to recordings from non-transgenic mice (Figure
2). These findings suggest altered network topology in APP23 limbic systems, which potentially include aberrant circuit activity of GABAergic interneurons and pyramidal cells. In human epilepsy and rodent models of epilepsy, pre-seizure states are characterized by enhanced gamma oscillations [
58]. However, these findings may not be directly translatable to epileptogenesis in APP transgenic mice. APP-expressing mice have been reported with changes in distribution and survival of GABAergic interneurons at early stages [
59]–[
61]. Cortical hypersynchronicity in APP mice and its relation to gamma oscillations and function of GABAergic interneurons has recently been investigated in detail [
34]. Verret
et al. elegantly showed that cortical gamma oscillations are impaired during strong hypersynchronicity in mutant APP transgenic J20 mice, compared to pre-ictal sequences within the same recordings. Different from this previous study, we focused on analysis of recording sequences that were not characterized by hypersynchronicity and made comparisons between APP23 and non-transgenic mice rather than of pre-ictal and ictal states. Thus, we do not conclude on gamma oscillations during hypersynchronicity states in the hippocampus of APP23 mice. Nevertheless, the study by Verret
et al., our study and findings by others [
36],[
54] indicate that altered network oscillations are an inherent phenotype of APP transgenic mice.
To our knowledge, our study represents the first report of hippocampal CFC deficits in adult APP transgenic mice. Previously, slice recordings have been analysed for theta phase to gamma amplitude coupling, however, only very young mice were used [
36]. Intriguingly, CFC and spectral changes were present before high levels of Aβ were detectable in these mice [
36]. These findings and our data suggest that CFC deficits occur early and persist over time in APP transgenic mice. They may thus contribute to hypersynchronicity and behavioural/memory deficits in APP transgenic mouse models that are already seen at this age. However, until now there are no mechanistic implications of CFC changes in epileptiform activity. So our study provides the first correlation of CFC deficits and hypersynchronous activity in the brains of APP-expressing mice. The above-mentioned study by Goutagny
et al also describes alterations in spectral contributions from theta and gamma oscillations in slices of APP transgenic TgCRND8 mice [
36], similar to what we detected in APP23 EEG recordings. CFC calculation measures used in our study (i.e. comodulogram plots, phase-amplitude distribution and modulation index) are based on normalized gamma amplitudes [
20]. Therefore, these measures allow for quantitative coupling strength comparison between APP23 and non-transgenic recordings.
Riluzole blocks persistent sodium currents by inhibiting voltage-gated sodium channels [
62].
In vivo, riluzole has anticonvulsive and sedative effects in rodents and humans [
63],[
64]. We found that riluzole injection in APP23 mice leads to strong induction of hypersynchronicity, an increase in theta oscillation power, yet a shift towards lower theta frequencies. It also enhanced the power of gamma oscillations. A synchronized hippocampal slow oscillation state lowers the threshold for propagation and generation of epileptiform activity [
57]. We found that riluzole causes frequency shift towards slow oscillations (≤4 Hz) in interictal EEG sequences, thus potentially instigating a slow oscillation state that lowers hypersynchronicity thresholds. This is neither seen with MK801 nor with SB203580, where interictal recordings still show a significant component of theta (4-12 Hz) in APP23 mice and induced hypersynchronicity was less pronounced (Table
1). Thus, the oscillatory state of hippocampal slow waves may be deterministic of seizure thresholds in APP transgenic mice.
Riluzole has been shown to enhance hypersynchronous states in cortical recordings from APP-expressing J20 mice [
34]. Interestingly, riluzole suppressed gamma oscillations in cortical recordings from J20 mice [
34], which is consistent with its effect on hippocampal gamma oscillations in APP23 mice in our study. We extended our analysis of riluzole-induced changes in EEG parameter to CFC between theta and gamma oscillations. Surprisingly, CFC strength was markedly increased in APP23 mice after riluzole injection, even though these mice show CFC impairment before injection. Even though non-transgenic mice show reductions in theta and gamma power, the effects of riluzole-injection on CFC in non-transgenic mice are insignificant, suggesting that riluzole has unique effects in APP-expressing mice that might impact directly or indirectly on generation and/or propagation of hypersynchronous activity. Our data are the first to correlate an increase in CFC strength with enhanced hypersynchronicity in EEG recordings in mice. The close correlation of riluzole-induced interictal surge in CFC strength and increased epileptiform activity in APP23 mice may imply a causal connection on a network level. Thus, our data using the APP23 mouse model tentatively imply that CFC analysis algorithms may prove useful to predict epileptiform episode in recordings of AD patients.
Voltage-gated sodium channels, targeted by riluzole, show altered expression and processing in mouse models of AD, which correlates with aberrant EEG activity and memory impairments, as well as altered sensitivity to riluzole [
34],[
60]. The contribution of riluzole-sensitive sodium channels in cortical interneurons has been suggested to underlie the hypersynchronous effect of riluzole on cortical neuron networks [
34]. This may to some extent translate to the hippocampal network in APP23 mice, as function of hippocampal GABAergic interneurons is impaired in epilepsy [
14]. In the rodent hippocampus, expression of riluzole-sensitive channels is not restricted to interneurons, but also occurs in pyramidal CA neurons [
65],[
66]. Thus, inhibition of voltage-gated sodium channels may have broad impact on neuronal networks in the hippocampus and – as secondary effect – may cause suppression of glutamate release [
63],[
67]. How, such broad sodium channel inhibition contributes mechanistically to hippocampal hypersynchronicity remains unclear. Interestingly, riluzole had opposite effects on theta and gamma spectral power in our recordings from non-transgenic and APP23 mice. This suggests that riluzole induces a specific interictal state in APP mice that combines increased low frequency theta, enhanced gamma oscillations and a pronounced gamma-theta coupling that might lower network thresholds for epileptogenesis. Given the direct mode of action of riluzole is inhibition of voltage-gated sodium channels, the dissimilarity of EEG responses of non-transgenic and APP23 mice is likely due to differences in expression of these channels as shown for other APP-expressing mice [
34].
Next, we used MK801, a non-competitive activity-dependent inhibitor of NMDA receptors, to address effects of NMDA receptor inhibition on hypersynchronicity and other EEG parameters in APP23 mice. Surprisingly, MK801 induced significant epileptiform activity in hippocampal recordings from APP23 mice, albeit to a lower extent as seen after administration of riluzole. Interestingly and different from the effects of riluzole, MK801 injection resulted in strongly enhanced theta oscillation and suppressed gamma oscillations, whilst not leading to frequency shifts (Table
1). MK801 also had significant effects of on theta and gamma spectral power in non-transgenic recordings. Therefore, we calculated the ratios of spectral power pre- and post-injection with MK801, which showed that theta, but not gamma, spectral power was significantly affected in APP23 mice. Furthermore, MK801 injection did not affect CFC strength. Thus, we argue that MK801 instigates a markedly different oscillatory state – in particular of theta rhythms - than riluzole in the hippocampus of APP23 mice, though both induce forms of hypersynchronicity. This suggests that a component in hypersynchronicity thresholds of APP23 mice is governed by NMDA receptors. Furthermore, neuronal populations may contribute differentially to hypersynchronicity in APP23 mice. In epilepsy models, MK801 inhibits epileptogenesis [
68] or has no effects [
69]. Therefore, the APP-dependent network alterations leading to hypersynchronicity may be unique to AD models with compromised NMDA receptor function. Interestingly,
stargazer mice with a mutation in the
Cacng2 gene that encodes stargazin, a protein that regulates AMPA receptor trafficking and excitability in interneurons, show increased seizure length and hyperexcitability upon NMDA receptor inhibition by MK801 [
70]. AMPA receptor trafficking may be affected in APP transgenic mice. Consistent with this idea, Aβ and expression of mutated APP have been shown to reduce synaptic prevalence of AMPA receptors, thereby affecting synapse function [
40],[
71].
We found p38 MAP kinase activity to be increased in APP23 mice. Interestingly, MK801, but not riluzole treatment results in lowered p38 MAP kinase activation in the hippocampus of APP23 mice (Figure
6A, Table
1). Surprisingly, suppressing hippocampal p38 activity with a pyridylimidazole-based inhibitor results in increased hypersynchronicity. Hence, increased p38 activity in APP23 hippocampus may be part of a protective mechanism limiting synchronized activity of neuronal circuits. MK801 treatment of APP23 mice resulted in similarly enhanced hypersynchronicity as after p38 inhibitor administration. Even though p38 inhibition – despite effective in the hippocampus – may affect other brain areas or organ systems (e.g. peripheral nervous system, vascular system), the effects seen with MK801 or p38 inhibitor could be mechanistically linked within hippocampal neurons. In support of this, NMDA receptors have been linked to p38 activation during promoting neuronal death [
43],[
49],[
72]. Interestingly, Aβ or mutated APP expression increases activation of p38 [
43],[
52]. Consistent with this, p38 might govern hypersynchronicity in APP mice downstream of MK801-sensitive NMDA receptors and this pathway may constitute a protective signal from epileptiform activity in this AD model. The molecular details of p38 signalling in this context as well as effects of long-term suppression of p38 activation on network aberrations and epileptiform activity remain to be investigated. Whether long-term inhibition of p38 is achievable in mice without affecting normal brain or peripheral function is, however, unclear, as longitudinal inhibitor studies of these functions are not available at present.
In summary, hippocampal EEG recordings in adult 4-month old APP23 mice show spontaneous hypersynchronicity, alterations in theta and gamma oscillation and impaired coupling of theta phase and gamma amplitude. Although, we found no apparent plaque burden in the 4 months-old APP23 mice analysed here, APP23 mice show memory deficits as early as 3 months of age and increased Aβ levels, as shown by others and us before [
42],[
44],[
55]. While we determined EEG alterations in APP23 mice at a single age (4 months), other APP-expressing strains show hypersynchronicity and network aberrations at earlier and later stages, suggesting that APP23 mice likely show similar changes at similar ages [
34],[
36],[
54]. Furthermore, differentially enhanced components of hypersynchronicity in APP23 mice after treatment with either riluzole, MK801 or SB203580, suggest involvement of sodium channels, NMDA receptors and p38 MAP kinase signalling in governing thresholds of hypersynchronicity in AD mouse models. The profound effects of short-term treatments with riluzole, MK801 or SB203580 on hippocampal EEG may well translate into changes in behaviour and memory performance specifically in APP23. Whether they are also reflected by changes in distribution or survival of neuronal subtypes (e.g. GABAergic interneurons) in the hippocampus of APP23 mice remains to be determined.
Competing interests
The authors declare that they have no competing interests.