Acute Effects of Two Different Species of Amyloid-β on Oscillatory Activity and Synaptic Plasticity in the Commissural CA3-CA1 Circuit of the Hippocampus

Recent evidence indicates that soluble amyloid-β (Aβ) species induce imbalances in excitatory and inhibitory transmission, resulting in neural network functional impairment and cognitive deficits during early stages of Alzheimer's disease (AD). To evaluate the in vivo effects of two soluble Aβ species (Aβ25-35 and Aβ1-40) on commissural CA3-to-CA1 (cCA3-to-CA1) synaptic transmission and plasticity, and CA1 oscillatory activity, we used acute intrahippocampal microinjections in adult anaesthetized male Wistar rats. Soluble Aβ microinjection increased cCA3-to-CA1 synaptic variability without significant changes in synaptic efficiency. High-frequency CA3 stimulation was rendered inefficient by soluble Aβ intrahippocampal injection to induce long-term potentiation and to enhance synaptic variability in CA1, contrasting with what was observed in vehicle-injected subjects. Although soluble Aβ microinjection significantly increased the relative power of γ-band and ripple oscillations and significantly shifted the average vector of θ-to-γ phase-amplitude coupling (PAC) in CA1, it prevented θ-to-γ PAC shift induced by high-frequency CA3 stimulation, opposite to what was observed in vehicle-injected animals. These results provide further evidence that soluble Aβ species induce synaptic dysfunction causing abnormal synaptic variability, impaired long-term plasticity, and deviant oscillatory activity, leading to network activity derailment in the hippocampus.


Introduction
Alzheimer's disease (AD), the most common type of dementia and progressive neurodegenerative disorder worldwide, is characterized by selective neuronal loss, and two histopathological features in postmortem tissue are extracellular amyloid plaques composed of amyloid beta peptide (Aβ) and intracellular neurofibrillary tangles composed of hyperphosphorylated tau protein [1]. Recent evidence indicates that soluble forms of Aβ induce glutamatergic, cholinergic, and GABAergic imbalance, resulting in functional impairment of neural networks during early AD stages [2][3][4][5]. In fact, Aβ-induced synaptic dysfunction precedes selective neuronal degeneration and may explain memory impairment during early AD stages and mild cognitive impairment, a prodromal stage of AD [6,7]. Although therapies based on modulation of GABAergic neurotransmission have been proposed for AD [8], current symptomatic therapies include cholinesterase inhibitors and NMDA antagonists only [9]. New Aβ-targeted immunotherapies have been tested in several clinical trials but without a clear clinical benefit [10]; therefore, no course-modifying treatment has been developed to date because of a lack of understanding of the fundamental mechanisms underlying AD, as well as the physiological role of amyloid peptides.
The hippocampus and entorhinal cortex are peculiarly susceptible to deleterious Aβ effects during early AD stages [21]. Hippocampal plasticity, necessary for learning and memory processes, is tuned by θ activity, which depends on acetylcholine release from the medial septum [2]. Moreover, the septum and hippocampus are reciprocally interconnected and functionally coupled through GABAergic and glutamatergic connections to form the septohippocampal system [2,22]. Each of these neurotransmitters contributes to hippocampal rhythmicity [23]. Moreover, θ and γ activities are associated through inhibitory synapses between GABAergic parvalbumin interneurons and pyramidal neurons [24]. Such oscillatory activity, including phase-to-amplitude coupling of θ and γ activity, is necessary for adequately encoding and storing information in the cortex and hippocampus [25][26][27]. In vivo studies have shown the relevance of the CA3-CA1 synapse in associative learning and memory processing [28,29] and its implications in AD through animal models [30]. In a very recent study, the electrophysiological activity of the CA3-CA1 region in humans was correlated with memory tasks (i.e., delayed match-to-sample) and it resembles synaptic hippocampal responses observed in rodents in the same areas [31]. Therefore, studies based on animal models may provide physiological information that could be applied in specific regions of clinical relevance in the human brain.
There are tight correlations among long-term potentiation (LTP) mechanisms, θ-γ oscillations, and hippocampaldependent memories [32][33][34]. Event-locked oscillatory activity in hippocampal formation and hippocampus-related structures is necessary for learning and long-term memory processes, as well as for declarative and spatial memory functions, which are impaired in early stages of AD [3,34,35]. Accumulating evidence indicates that Aβ affects θ, δ, and γ bands in different preclinical models of AD [3,36]. Similarly, EEG recordings in AD patients show pathological changes of network oscillations in a wide range of frequencies (i.e., α/θ ratio, γ coherence, θ and δ synchronization) [37,38]. Studies using an animal model of AD (transgenic CRND8 mice) indicate that alterations in θ-γ cross-frequency coupling might be used as an early biomarker of AD [39]. In another study, acute LTP impairment by Aβ 1-42 was related to alterations in oscillatory activity in θ-γ coupling at perforant path-dentate gyrus synapses [40]. Various studies have reported that changes in the spectral power of brain oscillations are related to LTP induction and expression [32,33,41,42] and have shown a relationship between changes in single-synapse and network oscillation activity. Therefore, Aβ-induced LTP impairment might be associated with oscillatory activity changes in brain structures affected during initial stages of AD.

Material and Methods
2.1. Ethical Statement. All procedures performed on living animals were performed in conformance with Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines [49], following the Guide for the Care and Use of Laboratory Animals (8th edition, National Institutes of Health) and fulfilling the Colombian regulation (Law 84/1989 and Resolution 8430/1993). In addition, every experimental design and all procedures were approved by the Universidad del Rosario Ethics Committee.

Animals.
Seventeen 16-20-week-old male Wistar rats, weighing 300 ± 30 g, were used as experimental subjects. Experimental animals were supplied by the Universidad Nacional de Colombia animal facilities. Animals were housed in a sound-attenuated room in polycarbonate cages, in groups of four, under controlled environmental conditions: 20 ± 1°C temperature, 50 ± 10% relative humidity, and 12 h light/dark cycle (lights on from 07:00 to 19:00). Animals had food and water available ad libitum. Experiments were performed in the morning. Special care was taken to minimize animal suffering and to reduce the number of animals used.

Neural Plasticity
Statistical power = 80%, and d = 14. The minimum sample number per group was 5. In brief, each experiment proceeded as follows: (1) under general anaesthesia, recording and stimulation electrodes were stereotactically inserted in CA1 and contralateral CA3; (2) once stable responses were obtained, the input/output relationship was established; (3) 30 min baseline recording was done to characterize basal synaptic efficiency; (4) the designed solution was microinjected in CA1; (5) the microinjected solution's effect on synaptic transmission was characterized by recording synaptic responses for 30 min after microinjection; (6) high-frequency tetanic stimulation was done to induce long-term plasticity; and (7) the effect of high-frequency stimulation (HFS) on synaptic efficiency was characterized by recording synaptic responses for 60 min after HFS (Figure 1(a)).

Surgery.
Under general anaesthesia induced with 1.5 g/kg of 25% urethane (Sigma-Aldrich, St. Louis, USA) and 10 mg/kg of xylazine (Rompun®, Bayer, Leverkusen, Germany) (intraperitoneal injection), the subject was placed in a rat stereotaxic frame (SR-6R, Narishige Inc., Tokyo, Japan). Nociceptive responses (tail prick and paw withdrawal reflex) were evaluated during the experiment. In case of slight motor response, an additional injection of urethane at 50% of the initial dose was administered. The respiratory and cardiac rates were monitored during the whole experiment, and a heating blanket was used to avoid hypothermia. Because we were interested in evaluating Aβ effects on NMDAdependent synaptic plasticity, we chose urethane as the anaesthetic agent, taking into account that it has been reported to have no significant effects on NMDA-type glutamate receptors when given in intermediate doses [51,52]; however, anaesthetic doses of urethane depress or even abolish 7-12 Hz atropine-resistant theta activity [53], but left 4-7 Hz atropine-sensitive theta oscillations unaltered. Longitudinal fronto-occipital incision, followed by connective and muscle tissue dissection, was used to expose the skull. Two holes were drilled in parietal bones: one for inserting a Diagram of a rat brain illustrating microelectrode cannula (for local field potential recording and microinjection in left CA1) and stimulating electrode (in right CA3) locations. (c) Panoramic photograph of an obliquely illuminated coronal brain slice containing representative electrolytic lesions made by passing current through recording (left) and stimulating (right) electrodes (upper panel), paired to a corresponding coronal diagram of the hippocampus (extracted and modified from the bregma: -3.6 mm diagram of Paxinos' rat brain atlas), summarizing recording (left, blue circles) and stimulating (right, red triangles) electrode placement in each experimental subject. (d) Panoramic photograph of a brain block containing methylene blue-coloured left hippocampus, attesting adequate diffusion of the injected solution.
3 Neural Plasticity recording electrode aimed at left CA1 (stereotaxic coordinates from the bregma: AP = −3:8 mm, L = 2:5 mm, left) and one for inserting a stimulating electrode aimed at right CA3 (stereotaxic coordinates from the bregma: AP = −3:8 mm, L = 3:7 mm, right) [54]. In order to reduce variability between subjects, we always use the same electrode configuration. The dura mater was cut and removed through the holes in order to allow electrode insertion (Figure 1(b)).

In Vivo Extracellular Electrophysiology.
A microelectrode cannula (a 5 MΩ impedance enamel-coated wire attached to a 25-G needle), for local field potential (LFP) recording and microinjection, was inserted 2.5 mm from the pial surface, through the skull hole for left CA1 (vide supra for stereotaxic coordinates), using a hydraulic micromanipulator (SM-25C, Narishige Inc., Tokyo, Japan). A stimulating concentric bipolar electrode was lowered 3.5 mm from the pial surface, through the skull hole for right CA3 (vide supra for stereotaxic coordinates), using a micromanipulator (SM-25A, Narishige Inc., Tokyo, Japan). A silver electrode was placed in neck musculature as a reference.
CA1 field activity was magnified (100x) using an ACcoupled preamplifier (NEX-1, Biomedical Engineering, New York, USA). The preamplified signal was then band-pass filtered (0.1 Hz and 10 kHz cut-off frequencies) and further amplified at 20x (yielding 2000x total gain). This conditioned signal was digitized using an analogue-to-digital converter (DigiData 1200, Molecular Devices, San José, USA) with 10 kHz sampling frequency and stored for offline analysis using commercial (Clampfit, Molecular Devices, San José, USA) and purpose-designed software.
CA3 was stimulated by applying 100 μs monophasic square pulses, delivered at 0.33 Hz frequency, using a stimulus isolation unit (Isolator-11, Molecular Devices, San José, USA), controlled by a pulse generator (9514 Plus, Quantum Composers, Bozeman, USA). Stimulus intensity (100-400 μA) was adjusted so as to obtain stable and reliable CA1 field excitatory postsynaptic potentials (fEPSP). The depth of the stimulating and recording electrodes from the pial surface was finely adjusted so as to obtain short latency fEPSP, with a waveform characteristic of CA1 stratum pyramidale.
Once a stable CA1 fEPSP was achieved, threshold stimulus intensity was established by decreasing the intensity to one-half the previous one until the fEPSP failure rate was equal to or higher than 50% (characteristically 50 to 200 μA). Stimulus intensity was then successively doubled until fEPSP response saturation was attained (characteristically 4 to 8 times the threshold intensity). Then, the stimulus intensity required to obtain~50% of maximal response (I 50 ) was selected to evaluate short-and long-term plasticity from then on. Basal synaptic activity was characterized by recording CA3 stimulation-evoked fEPSP in CA1 (I 50 intensity, 100 μs duration, and 0.1 Hz frequency) for 30 min. Intrahippocampal microinjection effect on basal synaptic activity was characterized by recording such fEPSP for 30 min after microinjection. Long-term potentiation at cCA3-to-CA1 synapses was induced by delivering six trains (1 s length, 100 Hz frequency) at 60 s intertrain intervals (HFS). HFS effect on cCA3-to-CA1 synaptic efficiency was studied by recording left CA1 response to right CA3 stimulation (I 50 intensity, 100 μs duration, and 0.1 Hz frequency) for 60 min. Deep anaesthetic level was maintained throughout the whole recording session using supplementary anaesthetic doses (about every 4 hours) to attain stable and reliable activity in CA1 [55].
Once the recording session ended, a terminal dose of anaesthesia (urethane 2 g/kg and xylazine 10 mg/kg, intraperitoneal) was given and the location of stimulating and recording electrode tips was marked by passing continuous current through them delivered by a precision current source (Midgard, Stoelting, Wood Dale, USA). The subject's brain was removed and submerged in 4% paraformaldehyde for 3 days; then, 100 μm thick coronal slices encompassing stimulating and recording sites were obtained using a vibratome (1000 Plus, Vibratome, Bannockburn, USA). The electrodes' position and diffusion trace of microinjection coloured solution were documented by examining the slices with a stereoscope (SZX16, Olympus, Tokyo, Japan) and taking digital photographs (Cybershot DSCW7, Sony, Tokyo, Japan) under oblique back-illumination (Figure 1(c)).
Aβ peptides have been repeatedly highlighted as crucial AD pathogenetic initiators. Although the underlying mechanism is not yet fully understood, some studies have indicated that Aβ can impair synaptic transmission and plasticity, leading to changes in spine density and, eventually, synaptic pruning [69][70][71]. We have found that intrahippocampal microinjection of soluble Aβ in anaesthetized rats affected cCA3-CA1 synapse variability and impaired long-term synaptic plasticity. Overall, such results concur with those of other studies, which have shown that high Aβ oligomer concentration interferes with synaptic efficiency and plasticity [61,62,72]. There is a mild increase in fEPSP slope variability after intrahippocampal injection of the vehicle. A similar effect has been observed in other types of in vivo preparations [73,74], possibly due to a mechanical and osmotic effect of the saline solution. Another possibility of that change may involve methylene blue; however, the toxic effects of this molecule in vitro have been reported at concentrations higher than 100 μM (around ten times our preparation) [75,76]. Methylene blue might inhibit Aβ oligomerization in a dosedependent manner (ranging from 0.01 to 445 μM), but that effect is observed only after several days of incubation (4 to 8 days) [77]. We found that injection of Aβ 1-40 , but not Aβ [25][26][27][28][29][30][31][32][33][34][35] , induces significant increases in variability in cCA3to-CA1 synaptic responses without significant changes in fEPSP slope. In agreement with our results, several in vivo studies found that the injection of different amyloid species did not affect hippocampal baseline synaptic potential amplitude or slope [60,74,78]. Synaptic variability is determined, among many factors, by presynaptic axonal noise, as well as release probability fluctuations [79]. On the one hand, Aβ  has been found to induce spike widening, which would increase synaptic release due to increased calcium influx into presynaptic boutons [80]; Aβ [25][26][27][28][29][30][31][32][33][34][35] has also been reported to produce spike broadening, but using doses one order of magnitude higher than the one used in our experiment [81]. On the other hand, Aβ 1-42 oligomers depress release probability at CA3-CA1 synapses [82]; moreover, glutamatergic synaptic transmission could be either enhanced or reduced by Aβ 1-40 , depending on its concentration and the specific pyramidal cell type affected [44]. Such opposing mechanisms could explain the observed Aβ 1-40 -induced synaptic variability increase, without significant changes in average fEPSP slope. It is plausible that the Aβ 25-35 concentration we used was not enough to induce significant changes in synaptic variability (however, it did affect other synaptic and network properties).
Although neither Aβ-intrahippocampal injection nor HFS induced significant CA1 global power spectrum changes (Suppl. Tables 3 and 4), specific frequency band rPSD computation evidenced that injection with Aβ 25-35 , but not Aβ 1-40 nor HFS, induced a significant γ and HFO 1 relative power increase. Transient (about 100 s) increases in γ and θ spectral power have been recorded in freely behaving adult rats' hippocampus immediately after LTP induction [32]. In the present experiment, such early PSD changes in control subjects were not detected given the 5 min window in our PSD analysis algorithm, which was not designed to detect short-lived changes.
Both γ oscillations and ripples (140-200 Hz) in the CA1 hippocampal region depend critically on the fast-spiking activity of parvalbumin-(PV-) expressing basket cells [83]. Aβ oligomers have been shown to directly interact with receptor tyrosine-protein kinase erbB-4, increasing its phosphorylation state [84]; erbB-4 activation increases PV interneuron-dependent oscillatory activity [85]. PV interneurons' increased activity, which has been described during early stages of AD models in association with subtle cognitive deficits [86,87], seems to represent an initial adaptive response to Aβ oligomer deposition, which is followed later by PV interneuron dysfunction and more severe cognitive deficits [7,88]. It is plausible that Aβ 25-35 has more affinity than Aβ 1-40 for erbB-4.
Both Aβ species used in our experiment (Aβ 25-35 and Aβ 1-40 ) impaired HFS-induced LTP in cCA3-to-CA1 synapses. This deleterious effect has been extensively reported, and many possible underlying mechanisms have been identified; among them, those pointing towards excitatory/inhibitory imbalance are relevant to our findings. Aβinduced calcium dyshomeostasis underlies distorted synaptic transmission, plasticity, and oscillatory activity in the brain. However, depending on exposure time, brain region, oligomer type, and receptor subunits involved, soluble amyloid has been reported not only to inhibit calcium 8 Neural Plasticity influx through NMDA receptors in cultured hippocampal [101] or cortical neurons [102] but also to increase NMDA-mediated calcium influx in mouse brains in vivo [103]. In addition, deleterious effects of Aβ 1-42 and Aβ 1-40 on NMDA function and LTP are reverted by a specific GLUN2B receptor antagonist [104]. Acute in vivo and ex vivo Aβ 1-42 administration deteriorates GABA B -mediated inhibitory transmission in CA3-to-CA1 synapses and impairs HFS-induced LTP of excitatory [48] as well as inhibitory potentials [105], and such effects are reverted by pharmacological activation of G-protein-gated inwardly rectifying potassium (GirK) channels. Aβ [25][26][27][28][29][30][31][32][33][34][35] has been found to act ex vivo as a GirK channel antagonist in CA3 pyramidal neurons [22]. Aβ-induced malfunction of NMDA and GirK channel conductance in pyramidal neurons might contribute, along with PV interneuron dysfunction, to hippocampal network instability, manifested through LTP impairment and aberrant rhythm generation, which may underlie subtle cognitive derailments observed during early AD stages. Intrahippocampal injection of Aβ 1-40 significantly shifted the resulting angle from the average θ-to-γ PAC vector. In vehicle-injected subjects, HFS induced a significant phase shift of the average θ-to-γ PAC vector persisting up to 60 min. Besides impairing HFS-induced LTP, both Aβ 25-35 and Aβ 1-40 blocked the phase shift of the average θ-to-γ PAC vector, with the Aβ 25-35 effect persisting longer. Increases in θ-to-γ coupling have been described in freely behaving rats after HFS-induced LTP; in that experiment, acute Aβ 1-42 treatment not only impaired LTP but also diminished θ-to-γ coupling [40]. θ-nested γ oscillations in CA1 depend on out-of-phase firing sequences of PV interneurons, pyramidal cells, somatostatin-positive (SST) neurons, and CA3-activated feedforward inhibitory interneurons during population θ oscillations, eventually opening windows for synaptic plasticity during specific θ phases [106]. Perisomatic inhibition by PV interneurons, associated with temporal silencing of feedforward inhibition acted by SST interneurons, allows calcium spike-associated plasticity; conversely, dendritic inhibition by feedforward interneurons prevents calcium spikes but facilitates pyramidal neuron output [106,107]. Phase synchronization of firing in such a network depends on the relative contribution of PV (γ band) and SST (θ band) interneurons [108]; therefore, θ-to-γ coupling phase shift represents specific variations in such contributions. The observed Aβ injectioninduced phase shift, as well as the impairment of HFSinduced phase shifts, may be due to its effect on PV interneurons [84,85,106] and on SST interneurons [108]. There may also be pyramidal cell contribution to such network dysfunction; in fact, Aβ 25-35 reduces GABA B -dependent GirK channel activity in pyramidal neurons [22]; this may enhance excitatory pyramidal cell influence on PV and SST interneurons, which would further destabilize the network. Interestingly, in a recent clinical study, restoration of temporal cortex θ-to-γ PAC was associated with working memory performance improvement in older adults [109].

Conclusions
The present study results indicate that changes in the functional relationships between θ, γ, and ripple oscillatory network activity in the hippocampus are highly correlated with amyloid-induced synaptic plasticity dysfunction in a model of early amyloid-β pathology (similar to what has been reported in early AD stages). Taken together, these results show that intrahippocampal microinjection of soluble forms of Aβ affects synaptic variability and plasticity and modifies neural processing and network activity, changes that might underlie cognitive deficits observed in early AD models. Aβ-induced derailment of the tight functional relationship in hippocampal circuits between θ oscillation (controlled by SST interneurons as well as by medial septum and entorhinal cortex inputs) and γ activity (controlled by PV interneurons) implies a dysregulation of the crosstalk of cholinergic, glutamatergic, and GABAergic systems during early AD stages, leading to impaired information processing and encoding. Therefore, the abnormality in θ-to-γ PAC hereby described is worth evaluating as a putative early biomarker of Aβinduced synaptic dysfunction in AD, long before neurodegeneration is established.
AD is a chronic and complex neurological disorder that involves several mechanisms (i.e., neuroinflammation and oxidative stress) additional to amyloid pathology. For that reason, it is important to start considering the role of θ-to-γ PAC in behavioural models of AD that involve tauopathy and selective chronic neurodegeneration, as well as to test it in further clinical trials through noninvasive electrophysiological methods in patients with mild cognitive impairment and major neurocognitive disorder.

Data Availability
Additional data set can be sent by request. Detailed information of methods and results are included in Supplementary Materials.

Disclosure
A preliminary version of this work was presented as a poster in the Joint Meeting of International Brain Research Organization & Federation of Asian-Oceanian Neuroscience Societies.

Conflicts of Interest
The authors declare that they have no conflicts of interest.

Authors' Contributions
Mauricio O. Nava-Mesa and Alejandro Múnera contributed equally to this work.