Background
Alzheimer’s disease (AD) is the most common age-related neurodegenerative disorder characterized by progressive memory loss and cognitive decline. Mutations in the
Presenilin genes account for ~90% of all causative mutations in familial AD, highlighting their importance in AD pathogenesis. Genetic studies using conditional gene targeting approaches revealed that Presenilins are essential for learning and memory, synaptic function and age-dependent neuronal survival [
1‐
3].
Synaptic dysfunction is widely thought to be one of the earliest key pathogenic events in AD before frank neurodegeneration [
1,
4‐
6], and the hippocampal network is particularly vulnerable in AD [
7‐
10]. The hippocampus consists of three main fields, dentate gyrus (DG), areas CA3 and CA1, and each field displays unique anatomical, molecular, and connectivity patterns [
11,
12]. The tri-synaptic circuit conducts synaptic transmission in the hippocampus, and consists of three major excitatory synaptic pathways: perforant path (PP) → DG, mossy fiber (MF) → CA3, and Schaffer collateral (SC) → CA1 [
13]. All three hippocampal pathways have been associated with learning and memory [
14‐
16], and disruption of the hippocampal network has been implicated in AD. For example, structural and functional MRI analysis of AD patients revealed disruption of the MF-CA3 pathway in patients with mild AD or mild cognitive impairment [
17,
18].
We previously reported that inactivation of Presenilins results in impairment of neurotransmitter release, short- and long-term synaptic plasticity at hippocampal SC synapses [
1,
2,
19]. However, it was unknown whether Presenilins play similar or distinct roles in the regulation of synaptic function at other hippocampal synapses. In the current study, we focus on the hippocampal MF pathway using the postnatal forebrain-restricted
Presenilin conditional double knockout (
PS cDKO) mice, in which Presenilins are inactivated in excitatory neurons of the hippocampus beginning at postnatal day 18 [
1,
2,
19,
20]. We found that long-term potentiation (LTP), paired-pulse facilitation (PPF) and synaptic facilitation are impaired at MF synapses in
PS cDKO mice. Moreover, post-tetanic potentiation (PTP), which lasts longer than facilitation and results from the slow efflux of tetanically accumulated mitochondrial Ca
2+ [
21,
22], is also reduced at MF synapses in
PS cDKO mice. Pharmacological blockade of mitochondrial Ca
2+ efflux mimics and occludes PTP deficits at MF synapses of
PS cDKO mice, suggesting an impairment of mitochondrial Ca
2+ at MF synapses in the absence of Presenilins. However, quantitative electron microscopy (EM) analysis showed similar numbers and areas of mitochondria between control and
PS cDKO mice at hippocampal MF presynaptic terminals. Consistent with these findings, Ca
2+ imaging of DG granule neurons showed that repeated stimulation-induced cytosolic Ca
2+ increases are impaired in granule neurons of
PS cDKO mice, and that blockade of mitochondrial Ca
2+ release mimics and occludes the Ca
2+ homeostasis deficits in
PS cDKO granule neurons. Taken together, our study demonstrates the importance of Presenilins in the regulation of synaptic plasticity and mitochondrial Ca
2+ homeostasis at hippocampal MF synapses.
Methods
Mice
The generation and extensive characterization of postnatal forebrain-restricted
PS conditional double knockout (
PS cDKO) mice were previously reported [
1,
2,
19,
20,
23]. Briefly, Northern, in situ hybridization and Western analyses were carried out to confirm the normal PS1 expression in
fPS1/fPS1 mice and the inactivation of PS1 in the cerebral cortex of
PS cDKO (
fPS1/fPS1; PS2
−/−
; αCaMKII-Cre) mice beginning at postnatal day ~18 and complete at ~4 weeks of age [
1,
2,
19,
20].
PS cDKO (
fPS1/fPS1; PS2
−/−
; αCaMKII-Cre) and control (
fPS1/fPS1) mice were in the B6/129 hybrid background. All electrophysiological, Ca
2+ imaging, quantitative EM and behavioral analyses were performed in a genotype blind manner using mice at the age of 2 months.
Preparation of brain slices
Hippocampal slices were prepared from both male and female PS cDKO and control mice at 2 months of age. Mice were decapitated after being anesthetized with ketamine (100 mg/kg) + xylazine (10 mg/kg) + acepromazine (3 mg/kg), and the whole brains rapidly removed and placed in ice-cold (4 °C) oxygenated (95% O2/5% CO2) high sucrose and magnesium solution containing (in mM) the following: 200 Sucrose, 25 NaHCO3, 10 Glucose, 3 KCl, 1.25 NaH2PO4, 1.2 Na-pyruvate and 0.4 Na-ascorbate, 7 MgCl2, and 0.5 CaCl2. Horizontal hippocampal slices (400 μm thick) were prepared using a vibratome (VT1200S, Leica, Germany), and transferred to an incubation chamber having oxygenated artificial cerebrospinal fluid (ACSF) containing (in mM) the following: 125 NaCl, 3 KCl, 1.25 NaH2PO4, 1 MgCl2, 2 CaCl2, 25 NaHCO3, 10 Glucose, 1.2 Na-pyruvate and 0.4 Na-ascorbate, adjusted to 310 ± 5 mOsm (pH 7.4). The slices were allowed to recover at 34 °C for 30 min and then placed in a recording chamber constantly perfused with heated ACSF (30 ± 1 °C) and gassed continuously with 95% O2 and 5% CO2. The flow rates of bathing solution and the volume of the recording chamber for slices were 2.2 ml/min and 1.2 ml, respectively. Hippocampal slices were visualized using an upright microscope equipped with differential interference contrast (DIC) optics (BX51WI, Olympus, Japan). The DIC optics was used for visualization of neurons in the course of whole-cell recordings. All experiment procedures were conducted in accordance with guidelines of the Brigham and Women's Hospital Institutional Animal Care and Use Committee and National Institutes of Health.
Electrophysiological analysis
For extracellular field recordings, stimulation pulses were delivered with a stimulus isolation unit (World Precision Instruments, A365) using a unipolar metal microelectrode. Stimulus electrodes were positioned ~600 μm from the recording electrode in the hilus adjacent to the DG granule cell layer (mossy fibers). Field excitatory postsynaptic potentials (fEPSPs) were recorded in current-clamp mode with ACSF-filled patch pipettes (1.5–2 MΩ). All fEPSPs were recorded with a stimulation strength that yielded 30% of the maximal response. To ensure that MF responses were not contaminated by associational/commissural inputs, the metabotropic glutamate receptor agonist (2S,1′R,2′R,3′R)-2-(2,3-dicarboxycyclopropyl) glycine (DCG IV; 2 μM) was applied at the end of recordings to block MF responses selectively. Data were included only if responses were reduced by more than 80%. All recordings were performed with the GABAA receptor antagonist bicuculline methiodide (10 μM) and NMDA receptor antagonist APV (50 μM) added to the ACSF. Data were collected with a MultiClamp 700B amplifier (Molecular Devices) and digitized at 10 kHz using the A/D converter DIGIDATA 1322A (Molecular Devices). Data were acquired and analyzed using a custom program written with Igor Pro software (Version 6.3; Wave-Metrics) and Clampfit (Version 10.3; Molecular device).
For input/output measurements, 10 traces were averaged for each stimulation intensity, and the amplitude of the fiber volley (FV) was measured relative to the slope of the fEPSP. The stimulation rate was 0.2 Hz. The average linear fit slope was calculated as the slope of the linear input/output relationship for each slice. In LTP recordings, after baseline responses were collected every 15 s for 15 min, LTP was induced by five episodes of theta burst stimulation (TBS) delivered at 0.1 Hz. Each episode contained ten stimulus trains (5 pulses at 100 Hz) delivered at 5 Hz. To generate summary graphs (mean ± SEM), individual experiments were normalized to the baseline, and four consecutive responses were averaged to generate 1 min bins. These were then averaged together to generate the final summary graphs. Paired-pulse facilitation (PPF) was measured as the ratio of the second fEPSP slope relative to the first fEPSP slope, evoked by two identical presynaptic stimuli. Synaptic facilitation was measured as the percentage of the fEPSP slope versus the first fEPSP slope at a given stimulus train in individual slices.
For whole-cell patch clamp experiments, recording pipettes (3–5 MΩ) were filled with a solution containing (in mM) the following: 120 K-gluconate, 10 KCl, 20 HEPES, 4 MgATP, 0.3 NaGTP, 10 phosphocreatine, and 0.1 EGTA with the pH adjusted to 7.30 with KOH (295–300 mOsm). Excitatory postsynaptic currents (EPSCs) at MF synapses were recorded from CA3-pyramidal cells (CA3-PCs) in voltage-clamp mode at a holding potential of −60 mV. The series resistance (Rs) after establishing whole-cell configuration was between 15 and 20 MΩ. Synaptic responses were evoked by extracellular stimulation via a stimulator (Stimulus Isolator A365; WPI) connected to a patch electrode filled with ACSF solution, and placed in stratum lucidum of CA3 field. The stimulus intensity was adjusted such that the baseline EPSC amplitude was in the range between 100 pA and 300 pA. After 10–15 min of stabilization from the break-in, EPSCs were evoked with 0.2 Hz stimulation and recorded for 3–5 min, and followed by high frequency stimulation (HFS: 16 pulses at 100 Hz, delivered 4 times at 0.33 Hz) to induce post-tetanic potentiation (PTP). The PTP was recorded in the presence or absence of CGP37157 (20 μM), inhibitors of mitochondrial Na+/Ca2+ exchanger (NCX), and CGP37157 treatment began 3 min before second HFS and lasted during the PTP recording. The magnitude of PTP was quantified as the average of the first three post-tetanic EPSC amplitudes normalized to the mean baseline amplitudes. EPSC recordings with >20% series resistance change were excluded from data analysis. At the end of each experiment, we examined the effect of DCG IV (2 μM) to confirm that we had studied MF synapses.
Quantitative EM
For quantitative EM analysis of mitochondria at MF synapses, four PS cDKO and four control mice at the age of 2 months were used. Animals were anaesthetized with sodium pentobarbital and perfused transcardially with physiological saline followed by fixative solution containing 1% glutaraldehyde and 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). Fixed brains were isolated and stored in fixative solution at 4 °C overnight, washed in phosphate buffer, and sectioned coronally on a vibratome (Leica VT 10005) at a thickness of 200 μm. Dorsal hippocampi were carefully excised and post-fixed in 1% OsO4 for 30 min. After rinsing in distilled water and dehydration in an ascending series of ethanol (block-staining with 0.5% uranyl acetate in 70% ethanol) followed by propylene oxide, the hippocampi were embedded in Epon (Fluka) and hardened at 65 °C for two days. Thin sections from stratum lucidum of area CA3 were cut on an ultratome (Leica Ultracut) and mounted on formvar-coated 50-mesh copper grids. Sections were post-stained with lead citrate and subjected to electron microscopy (6200× magnification). The number and area of mitochondria and the area of the presynaptic bouton profiles were quantified. To avoid multiple measurements of the same bouton, randomized sections 5 μm apart from each other were analyzed and at least 10 micrographs per mouse in the cohorts of four animals per condition were used.
Calcium imaging
Cytosolic Ca2+ ([Ca2+]i) were measured in somata of hippocampal DG granule cells (GCs) using 100 μM Fura-2 pentapotassium salt introduced by a patch pipette. Ca2+ transients were evoked by 10 repetitive depolarizing pulses (2 ms duration from −80 to 0 mV) at various frequencies (1, 5, 10, and 20 Hz) under voltage-clamp conditions. For PTP induction, Ca2+ transients were elicited by 16 repetitive depolarizing pulses at 100 Hz, delivered 4 times at 0.33 Hz. Fluorescence imaging was performed with a 20× water-immersion objective (NA 0.5, UMPlanFln, Olympus), an air-cooling digital monochrome interline CCD camera (ORCA R2, Hamamatsu) and a monochromator (Polychrome V, FEI), which were controlled by a computer and Control Unit for real time TILL Imaging (FEI), running a Live Acqusition Imaging software. Images were taken at 20 Hz with double wavelength excitation at 340 and 380 nm. The ratio r = F
340/F
380 was converted to [Ca2+] values. Calibration parameters were determined using the in vitro calibration method (Invitrogen calibration kit). R
min and R
max values were calculated as 0.200 and 2.207, respectively.
Behavioral analysis
Mice were housed in a standard 12 h light–dark cycle. Six PS cDKO and six control male mice at 2 months of age were used in the Morris water maze behavioral test. Mice were handled daily for 5 days before training or testing. The Morris water maze is a circular pool 160 cm in diameter. During water maze tasks, mice were trained for 13 days with four trials daily, and the probe tests were administered at days 7 and 13. Mice were released from all four quadrants in a pseudorandom manner during the four training trials. Four visible board cues were hung on the walls during training and probe tests. During the hidden platform training, the platform (10 cm in diameter) was kept submerged under water and maintained in the same position. Each mouse was given four trials daily with a maximum duration of 90 s separated by a minimum of 15 min. If mice did not find the hidden platform, they were guided to the platform and allowed to remain on it for 15 s. The swimming of the mice was monitored using an automated tracking system (HVS Image). After 6 days and 12 days training, two probe tests were performed in the morning of day 7 and day 13 before new training trials. During the probe test at day 7 and day 13, the hidden platform was removed and a 90 s probe test was performed. The mice were released from the opposite position of the platform and searching for the location of the hidden platform according to four visible board cues on the walls. After the 90 s probe test, mice were taken out of the pool and the platform was re-installed in the same position followed by four new training trails.
Statistical analysis
Statistical analysis was performed using two-tailed Student’s t-test or two-way repeated-measures ANOVA with Bonferroni correction to test for significance for all comparisons of the electrophysiological, quantitative EM analysis, Ca2+ imaging and behavioral results. All data are presented as mean ± SEM.
Discussion
The hippocampus is known to be particularly vulnerable in AD, and is composed of three major circuits [
7‐
9,
17,
18]. Our previous studies of Presenilins and Nicastrin, another essential component of the γ-secretase complex, in synaptic function, however, focused exclusively on the hippocampal SC pathway [
1,
2,
19,
34]. In the current study, we investigate the role of Presenilins at hippocampal MF synapses to determine whether Presenilins employ a universal or distinct mechanism to control synaptic function in the hippocampus. Our electrophysiological, quantitative EM and imaging analyses revealed the essential role of Presenilins in the regulation of synaptic plasticity and mitochondrial Ca
2+ homeostasis at hippocampal MF synapses.
Similar to SC synapses, we found that presynaptic short-term plasticity, such as PPF and synaptic facilitation, and LTP are impaired at hippocampal MF synapses in the absence of Presenilins (Figs.
1 and
2), indicating a universal requirement of Presenilins for normal synaptic plasticity at the hippocampal SC and MF synapses. These findings are consistent with the spatial learning and memory deficits exhibited by these
PS cDKO mice at 2 months of age in the hippocampal memory-dependent Morris water maze task using either an intensive training protocol (6 trials per day) in a prior study [
1] or a less intensive and more difficult training protocol (4 trials per day) in the current study, which revealed more dramatic learning and memory deficits (Fig.
7). Interestingly,
PS1 cKO mice also exhibited mild but significant learning and memory deficits using a similarly difficult training protocol, which was designed to uncover more readily spatial learning and memory impairment [
20]. Thus, the severity of learning and memory deficits observed in the water maze is Presenilin dose dependent with
PS cDKO mice exhibit more severe phenotypes than
PS1 cKO mice. The synaptic plasticity impairments observed in the MF pathway of
PS cDKO mice in the current study and previously reported in the SC pathway [
1,
2,
19] likely contribute to the spatial learning and memory deficits identified in the current and prior studies (Fig.
7, [
1]). For example, MF synaptic plasticity was reported to be important for the establishment of hippocampus-dependent associative learning and spatial memory [
35‐
38]. Furthermore, spatial learning and memory analyzed in the water maze results from network interactions between hippocampal tri-synaptic circuits and the entorhinal cortex. At 2 months of age, postnatal forebrain-restricted
PS cDKO mice exhibit normal numbers of cortical and hippocampal neurons as well as normal volume of the neocortex and hippocampus, indicating unaffected cortical development in these mice [
1,
3], despite neurodevelopment phenotypes observed in
PS germ-line mutant mice and neural progenitor cell lineage restricted conditional mutant mice [
39‐
42]. Future studies will be needed to determine whether and how Presenilins control synaptic plasticity in the hippocampal perforant path (PP).
The impairment in the LTP induction phase is more dramatic at MF synapses of
PS cDKO mice (Fig.
1) than what was previously reported at SC synapses [
2,
43]. This is likely due to the fact that high-frequency stimulation at MF synapses induces multiple forms of synaptic strength enhancements, including PTP, in LTP induction phase [
44,
45]. Furthermore, in contrast to the SC and the perforant path, MF synapses display a particular form of LTP that is mainly expressed presynaptically, and is independent of NMDA receptor activation [
45‐
47]. The early induction phase of LTP at MF synapses is triggered by a tetanus-induced rise in presynaptic intracellular Ca
2+, which results in activation of a Ca
2+/calmodulin-activated adenylyl cyclase [
47‐
50]. The mitochondrial Ca
2+ deficit observed in presynaptic neurons of the MF pathway in the absence of Presenilins (Fig.
6) likely underlies the greater impairment of early phase LTP induction at MF synapses relative to SC synapses.
Interestingly, we found that another form of presynaptic short-term plasticity, PTP, is impaired at MF synapses in
PS cDKO mice (Fig.
3). Due to their unique structural features, CA3 pyramidal neurons receive excitatory synapses from stellate cells of layer II of the entorhinal cortex onto their distal apical dendrite [
51,
52], and from other CA3 axon collaterals onto the remainder of the apical and the entire basal dendrite [
53]; thus, the MF-CA3 projection may be contaminated with polysynaptic responses. We therefore performed whole-cell patch recording to ensure that the EPSCs recorded at MF synapses were monosynaptic. The latter is supported by the observations that the EPSC’s rise times are uniform, and their latencies are relatively short, and their distribution is unimodal (Fig.
3). Furthermore, if our EPSC recording were contaminated with polysynaptic contributions, then increasing the intensity of presynaptic stimulation would expect to result in slower-decaying synaptic currents. However, we found that the EPSCs decay time constant (τ) did not correlate with the EPSC amplitude or stimulation intensity (Fig.
3). This is consistent with monosynaptic nature of the recorded EPSCs [
27,
28], and suggests that our EPSC recording in the MF pathway reflected monosynaptic responses with no significant contamination by polysynaptic components.
PTP is known to be dependent on mitochondrial Ca
2+ and is longer lasting than frequency facilitation due to the slower release of Ca
2+ from mitochondria [
21,
22,
25,
54]. Indeed, blockade of mitochondrial Ca
2+ release by NCX inhibitor CGP37157 mimics and occludes the PTP impairment observed at MF synapses of
PS cDKO mice (Fig.
3), indicating that the PTP deficits in
PS cDKO mice are due to the mitochondrial Ca
2+ defects. However, quantitative EM analysis revealed similar number and area of mitochondria at presynaptic boutons of control and
PS cDKO MF synapses (Fig.
4). Ca
2+ imaging analysis of acute hippocampal slices demonstrated that Presenilins are essential for normal mitochondrial Ca
2+ homeostasis at MF synapses (Figs.
5 and
6). We measured the ∆[Ca
2+]
i increments in the cell body of DG granule neurons instead of presynaptic axon terminals, and the cytosolic Ca
2+ increases are reduced in DG granule neurons of
PS cDKO mice when induced by tetanic stimulation between 5 and 20 Hz but unchanged at 1 Hz (Fig.
5). However, synaptic facilitation induced by repeated stimulation at 1, 5, 10 and 20 Hz is impaired at MF synapses of
PS cDKO mice (Fig.
2). The difference between cytosolic Ca
2+ increases and synaptic facilitation induced by repeated stimulation at 1 Hz is likely due to the fact that during stimulation [Ca
2+]
i increments are much higher in presynaptic axon terminals and have faster kinetics, compared to cell bodies, because of their different Ca
2+ clearance mechanisms and endogenous Ca
2+ buffers [
55‐
60].
Blockade of mitochondrial Ca
2+ release mimics the impairment of cytosolic Ca
2+ increases elicited by PTP induction stimuli (16 pulses at 100 Hz, delivered 4 times) in DG granule neurons of
PS cDKO mice (Fig.
6). Furthermore, the amplitude of cytosolic Ca
2+ increases elicited by high frequency stimulation (>20 Hz) is similarly reduced in DG granule neurons of
PS cDKO slices and in DG granule neurons of control slices treated with CGP37157, whereas CGP37157 has little effect in
PS cDKO DG neurons (Fig.
6). These results suggest that the mitochondrial Ca
2+ deficits likely contribute to the presynaptic impairment observed at MF synapses of
PS cDKO mice. How Presenilins control mitochondrial Ca
2+ homeostasis is unknown. We previously reported that ryanodine receptor (RyR)-mediated Ca
2+ release from the ER is impaired in the absence of Presenilins [
2,
61]. Furthermore, RyR levels are reduced in the hippocampus of
PS cDKO mice but IP
3 receptors and SERCA are unchanged [
61]. It remains to be determined whether Presenilin regulates mitochondrial homeostasis via its uniporter and/or antiporters or through its modulation of Ca
2+ release from the ER, since communication between ER and mitochondrial membranes is thought to facilitate Ca
2+ transfer [
62‐
64]. Since mitochondrial Ca
2+ dysregulation likely contributes to apoptotic neuronal death observed in
PS cDKO mice during aging [
3], future studies will aim at elucidation of the molecular mechanism by which Presenilins control mitochondrial Ca
2+ homeostasis, which may be explored to prevent neurodegeneration caused by Presenilin dysfunction.