Abstract
Exercise is becoming a promising therapeutic approach to prevent alterations both in Alzheimer’s disease (AD) patients and in transgenic models of AD. This neuroprotection has been associated with changes in hippocampal structure and function, as well as with the reduction of amyloid-β (Aβ) production and accumulation. However, whether exercise produces lasting changes in hippocampal population activity and renders it resistant to Aβ-induced network dysfunction is still unknown. Thus, we tested whether voluntary exercise changes hippocampal population activity and prevents its alteration in the presence of Aβ, which has been associated to glycogen synthase kinase-3β (GSK3β) activation. We found that the hippocampal population activity recorded in slices obtained from mice that exercised voluntarily (with free access to a running wheel for 21 days) exhibits higher power and faster frequency composition than slices obtained from sedentary animals. Moreover, the hippocampal network of mice that exercised becomes insensitive to Aβ-induced inhibition of spontaneous population activity. This protective effect correlates with the inability of Aβ to activate GSK3β, is mimicked by GSK3β inhibition with SB126763 (in slices obtained from sedentary mice), and is abolished by the inhibition of PI3K with LY294002 (in slices obtained from mice that exercised). We conclude that voluntary exercise produces a lasting protective state in the hippocampus, maintained in hippocampal slices by a PI3K-dependent mechanism that precludes its functional disruption in the presence of Aβ by avoiding GSK3β activation.
INTRODUCTION
Amyloid-β (Aβ) plays a major role in cognitive deficits that characterize Alzheimer’s disease (AD) [1–5], which are mainly mediated by Aβ-induced neural-network dysfunctions [3, 7]. We have previously shown that Aβ decreases the power of hippocampal-entorhinal cortex population activity both in vivo [8, 9] and in vitro [10] by activating a transduction pathway that involves several protein kinases such as Fyn [5, 11] and glycogen synthase kinase-3β (GSK3β) [7–9]. In fact, either the absence or the pharmacological inhibition of any one of these kinases precludes the Aβ-induced reduction in hippocampus-entorhinal cortex population activity [5, 12]. These findings provide mechanistic support for the protective effect of GSK3β inhibition on the cognitive impairment induced by Aβ application [13], that is observed in transgenic AD models too [14].
Interestingly, there are non-pharmacological approaches that also reduce GSK3β activity [15–17]. Among these, exercise not only inhibits GSK3β [15, 16] but also improves hippocampal function and cognition in normal humans [18] and AD patients [4, 20], as well as in naïve [21–24] and AD-transgenic rodents [25]. Thus, while acknowledging that exercise can reduce Aβ production and accumulation [26], we speculate that the beneficial cognitive effects of exercise in AD [4, 20] and AD-animal models [27] could involve changes in the hippocampal network function [23] and Aβ-induced GSK3β activation [5, 10], which would protect the hippocampus against Aβ-induced inhibition of neural network activity [9, 28]. Thus, we aimed to test whether or not voluntary exercise modifies hippocampal activity and the activation of GSK3β by Aβ. We found that voluntary exercise changes the hippocampal population activity, which also becomes insensitive to Aβ-induced inhibition. This protective effect correlates with the inability of Aβ to activate GSK3β, possibly involving phosphoinositide-3-kinase (PI3K).
MATERIALS AND METHODS
Animals
All experimental procedures were approved by the Bioethics Committee of the Instituto de Neurobiología, UNAM, and were carried out according the Norma Oficial Mexicana de la Secretaría de Agricultura (SAGARPA NOM-062-ZOO-1999), which complies with the guidelines of the Institutional Animal Care and Use Committee Guidebook (NIH publication 80-23, Bethesda, MD, USA, 1996). Five-week-old male CD1-mice were maintained in a vivarium and divided into two groups: Sedentary-group (SED) animals were housed individually for three weeks in the same cages where they bred. Simultaneously, exercised-group (EX) animals were housed individually in cages with free access to a running wheel [29]. Animals were left undisturbed in these conditions for 21 days before any further experimental procedure.
Slice preparation
To obtain hippocampal slices, both SED and EX animals were anesthetized with an intraperitoneal injection of sodium pentobarbital (63 mg/Kg) and perfused transcardially with a cold, protective-saline solution containing 238 mM sucrose, 3 mM KCl, 2.5 mM MgCl2, 25 mM NaHCO3, and 30 mM D-glucose, pH 7.4, and bubbled with carbogen (95% O2 and 5% CO2). Then, animals were decapitated; their brains were removed and dissected in ice-cold artificial cerebrospinal fluid (aCSF) containing 119 mM NaCl, 3 mM KCl, 1.5 mM CaCl2, 1 mM MgCl2, 25 mM NaHCO3, and 30 mM D-glucose, pH 7.4, and bubbled with carbogen. One cerebral hemisphere was mounted onto an agar block with a 30° inclination, and horizontal 400-μm-thick slices containing the hippocampus were cut with a vibratome (Vibratome, St. Louis, MO, USA [2]). Slices were allowed to recover in aCSF at room temperature for at least 60 min before any further experimental manipulation.
Spontaneous population activity recordings
After of the recovery period, the hippocampal slices were transferred into a recording chamber continuously perfused in aCSF recirculating at 17-20 ml/min and 29±1°C. An extracellular electrode (0.5-1.0 MOhms) was located, using a stereoscopic microscope and a micromanipulator, on the CA1 region to record its spontaneous population activity (CA1SPA). The signal was amplified and slightly filtered (highpass, 0.5 Hz; lowpass, 2.5 KHz) with a wide-band AC amplifier (Grass Instruments, Quincy, MA, USA). This raw signal represents the activity of a local network of neurons [8, 30]. Considering that most of its power (Fig. 1) is concentrated in low-frequency components (0-50 Hz), it might be considered a local field potential (normally filtered at < 500 Hz), which is believed to be generated by membrane currents of the neurons in the local neighborhood of the recording electrode [30]. However, considering that the activity is obtained from the CA1 pyramidal cell layer and that we use a 2.5 KHz lowpass filter, the activity also contains the spiking of local neurons [8, 30]. For these reasons, we call this signal CA1SPA and consider that it represents the alternating expression of different population activity patterns (see [31]). Electrodes were left in position for at least 15 min before commencing recordings. Control experiments were performed to confirm that CA1SPA remains stable for at least 3 h (Fig. 1; [6]). In all cases basal activity was recorded for 30 min. In some experiments, Aβ 10 nM was added to the bath after basal recordings, and its effects were continuously recorded for 60 min. In another set of experiments, we evaluated the effect of Aβ 10 nM in the presence of either the GSK3β-specific inhibitor SB126763 10 nM [12] or the PI3K inhibitor LY294002 10μM [32], which where bath applied after basal recordings, and its effects were continuously recorded over 60 min. Then, in the continuous presence of either kinase inhibitor, Aβ 10 nM was added to the bath, and its effects were continuously recorded for another 60 min. At the end of all experiments, lidocaine 1 mM was bath applied to verify the viability of the slice [10]. The signals were digitized (at 3 KHz) and stored in a personal computer with an acquisition system from National Instruments (Austin, TX, USA) using custom-made software designed in the LabView environment.
Aβ preparation
Aβ1 - 42 was obtained from American Peptides. The oligomerization protocol was performed as previously described [10, 33]. Briefly, 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP) was added to solid Aβ1 - 42, at a final peptide concentration of 1 mM, and incubated for 60 min at room temperature. HFIP was allowed to evaporate overnight. Then a 5 mM solution was prepared by adding DMSO. Such solution was then diluted with F12 medium to reach a final concentration of 100μM. This solution was incubated at 5°C for 24 h. Finally, this solution was centrifuged at 14,000 rpm for 10 min in the cold, and the supernatant containing the Aβ oligomers, as well as monomers [10], was collected and used for the experiments.
Western blot
For western blot analysis of the GSK3β phosphorylation-state, we use independent slices that underwent experimental procedures previously described [12, 34], after which their protein content was extracted [12, 34]. Slices were placed in a frozen tube and homogenized in ice-cold lysis buffer, containing: 50 mM Trizma Base, 150 mM NaCl, 1% Triton, 0.5% SDS, 25 mM NaF, 1 mM Na4P2O7, 10 mM Na3VO4, 20 mM β-glycerophosphate, and protease inhibitors [12, 34]. Samples were pre-cleaned by centrifugation at 14,000 rpm and stored at –80°C. For electrophoresis, cell homogenates were boiled for 10 min and centrifuged for 5 min at 10,000 rpm. After adding 2X Laemmli sample buffer, 15μL per lane of the total homogenate was loaded onto a 10% SDS-PAGE gel. After electrophoresis, proteins were transferred onto a polyvinyl difluoride (PVDF) membrane and were incubated for 1 h in a blocking buffer consisting of 20 mM Trizma base, 137 mM NaCl, and 0.05% Tween (TBS-T), pH 7.6 and 7.5% nonfat dry milk. After that, the membranes were incubated with phospho-GSK-3α/β (Ser21/9) rabbit polyclonal antibody (Cell Signaling Technology) diluted in TBS-T buffer at 4°C. The membranes were then washed twice with TBS-T and twice with normal TBS (20 mM Trizma base and 137 mM NaCl, pH 7.6), then incubated with the horseradish peroxidase-conjugated secondary antibody (goat anti-rabbit IgG-HRP Ab; SantaCruz Biotechnology) at a 1:5,000 dilution in TBS-T for 1–2 h. Proteins were visualized using Western Blot Luminol Reagent (Santa Cruz Biotechnology). The membranes then were reanalyzed for total GSK3β using the GSK3β rabbit monoclonal antibody (Cell Signaling Technology). To do so, PVDF membranes were stripped in a buffer solution (containing 0.070 mL 2-mercaptoethanol, 2 mL 10% SDS, 1.26 mL 0.5 M Tris, pH 6.8, and 6.67 mL deionized water) for 30 min at 50°C with occasional agitation. Thereafter, the membranes were washed four times with deionized water, twice with normal TBS, and then processed as described above.
Data analysis
Segments of 40 s of the recording, at the end of each experimental condition, were analyzed by means of a Rapid Fourier Transform, with a Hamming window, using Clampfit (Molecular Devices). The power between 1-50 Hz was integrated for each slice and each condition. We also used the time-frequency analysis available in EEGLAB (http://sccn.ucsd.edu/eeglab.
The integrated power of the activity of SED and EX slices was compared with an unpaired t-test. The integrated power after Aβ application was compared against control activity using a paired t-test. To analyze the difference between the activity of SED and EX slices in control conditions, we first calculated the relative power as the ratio of each frequency power to the total power between 1-50 Hz. To compare the difference of relative power, a Mann-Whitney test was performed for each frequency. We also performed a spectral difference analysis to represent the % change for the power of each frequency in the EX-group spectra with respect to the mean power of the same frequency in the SED-group spectra. To compare the differences between the SED and EX-groups, a Mann-Whitney test was performed for each frequency. Photographs of western blots were scanned, and the optical density (OD) was quantified with ImageJ. To obtain the GSK3β phosphorylation ratio, the OD of p-S9GSK3β was divided by the OD of total GSK3β for all the conditions. To compare the differences of pGSK3β/tGSK3β before and after Aβ application we use a paired t-test. To compare the CA1SPA the GSK3β phosphorylation ratio after the treatment with SB126763 10 nM or LY294002 10μM and with the corresponding values after the subsequent application of Aβ 10 nM, the activity was normalized to the control conditions (set as 100%). To compare the differences between these groups, a Friedman test with a Dunn’s post hoc test were performed. The statistical analysis was performed with Prism software, and a p value < 0.05 was considered significant.
RESULTS
CA1 spontaneous population activity and GSK3β phosphorylation are stable over time in vitro
As previously reported, CA1 spontaneous population activity (CA1SPA) in vitro is characterized by low-voltage, non-rhythmic activity that includes a broad range of frequency components whose power is stable for several hours [8, 10] (Fig. 1A). Similarly, in this study we found that CA1SPA power (quantified as % of control, i.e., basal recording) does not change after continuous recording for 60 min (104.40±5.26% of control), 90 min (100.30±7.16% of control), or 150 min (98.39±19.14% of control) (Fig. 1A) (n = 7). Since we have previously shown that GSK3β is involved in the inhibition of spontaneous population activity induced by Aβ [12], we also tested the stability of GSK3β activity (measured as serine phosphorylation [12]) in our experimental conditions. Indeed, we found that GSK3β phosphorylation does not change after the slices are continuously present in the bath for 60 min (94.74±28.19 % of control), 90 min (114.30±14.27% of control), or 150 min (109.90±8.01% of control) (n = 3).
Exercise changes CA1 spontaneous population activity
Before testing the effects of Aβ on CA1SPA we tested whether exercise changes this activity (Figs.1 and 2). We found that the CA1SPA showed higher integrated power (from 1 to 50 Hz) in slices obtained from the EX than from the SED group. The integrated power of the CA1SPA obtained from the EX slices was 4.12±0.58μV2, which is significantly higher (p < 0.05, n = 7) than the power found in slices obtained from the SED group (2.53±1.62μV2, n = 7) (Fig. 2B). Exercise not only enhances the power of the CA1SPA of slices obtained from the EX group, but it also changes its frequency components (Fig. 1). For instance, the CA1SPA of EX group slices exhibits enhanced power in the frequency band between 16 and 20 Hz, whereas its power is reduced around 4.8 Hz compared with the CA1SPA of slices obtained from the SED group (Fig. 1B-D, n = 7, Mann-Whitney test).
Exercise protects CA1 spontaneous population activity against Aβ-induced inhibition
After recording the CA1SPA in control conditions in slices obtained from the EX and SED groups, Aβ was added to the recording chamber, and its effect was evaluated for 60 min. As previously shown [10, 12], Aβ produces a generalized decrease in the power of all frequency components of the CA1SPA, which reduces the integrated power from 2.53±1.62μV2 to 1.50±0.73μV2 (a reduction of 50.01±12.27%, Fig. 2B, n = 7, paired t-test). This reduction in the integrated power of the CA1SPA of slices obtained from the SED group closely correlates with the reduced serine phosphorylation of GSK3β, which is indicative of its activation, passing from 0.37±0.07 O.D. to 0.25±0.07 O.D. (Fig. 2 C, n = 5, paired t-test). In contrast, when Aβ was applied to slices obtained from the EX group, the integrated power of the CA1SPA remained without change at 4.08±0.61μV2 (Fig. 2B, n = 7). This resistance of the CA1SPA of slices obtained from the EX group to the inhibitory effect of Aβ also correlates with the inability of Aβ to alter the phosphorylation state of GSK3β, which remains unchanged when comparing its levels in control conditions (0.41±0.09 pGSK3β/tGSK3β) to those after 60 min in the continuous presence of Aβ (0.51±0.04 pGSK3β/tGSK3β) (Fig. 2C, n = 5).
The inhibition of CA1 spontaneous population activity induced by Aβ depends on GSK3β activation
As previously demonstrated in the entorhinal cortex [12], the inhibitory effects of Aβ on spontaneous network activity require GSK3β activity. We tested whether this mechanism is also involved in the inhibition of CA1SPA observed in slices obtained from the SED group as follows: after recording control activity, the specific GSK3β inhibitor SB126763 10 nM was bath applied, and its effect was evaluated for 1 h. Then, Aβ was applied, and its effect was evaluated for 1 h in the continuous presence of SB126763 (Fig. 3). The power of the CA1SPA of slices obtained from the SED groups remained unchanged after the application of SB126763 10 nM (114.60±14.92% of control). Moreover, the power of CA1SPA also remained unchanged after Aβ application in the continuous presence of SB126763 (118.88±13.53% of control, n = 7) (Fig. 3). The western blot analysis of slices obtained from the SED group showed that the GSK3β phosphorylation state did not change after the application of SB126763 10 nM (109.80±24.94%, n = 3) or after the subsequent application of Aβ in the continuous presence of SB126763 (105.70±20.90%, n = 3).
The protective effect of exercise on CA1 spontaneous population activity against Aβ-induced inhibition requires PI3K activity
Considering that several of the protective effects of exercise require the activation of PI3K [35, 36], we tested whether this was the case for the protection of CA1 spontaneous population activity against Aβ-induced inhibition. Thus, we tested the effect of the specific PI3K inhibitor LY294002 10μM on the CA1SPA of slices obtained from the EX group and whether this inhibitor altered the effects of Aβ. After recording the control activity of slices obtained from the EX group, LY294002 10μM was applied and recording was continued for 1 h. Then, Aβ was applied and its effect was evaluated for one hour in the continuous presence of LY294002 (Fig. 4). After the application of LY294002 10μM, the power of CA1SPA remained unchanged (105.40±14.92% of control), but upon Aβ application in the continuous presence of LY294002, the power of the CASPA decreased significantly to 60.37±13.50% of control (Fig. 4, n = 7, Friedman test). The western blot analysis of slices obtained from the EX group and subjected to the same experimental procedure showed that the inhibition of PI3K is not sufficient to change the phosphorylation state of GSK3β (115.70±31.82% of control), but that upon Aβ application, there is a reduction in GSK3β phosphorylation to 73.77±12.27% of control (Fig. 4, n = 4, Friedman test with Dunn’s post hoc test).
DISCUSSION
Our results show that 3 weeks of voluntary exercise induce a persistent modulation of the hippocampal circuit that changes its CA1SPA (Fig. 1B-C) and makes it resistant to the deleterious effects of Aβ ex vivo by avoiding GSK3β activation and by promoting PI3K activity (Fig. 5). Furthermore, our results show a strong correlation between Aβ-induced CA1SPA inhibition and GSK3β activation (Fig. 5). Although the source of the reduction in spontaneous population activity induced by Aβ has not been precisely revealed, there is evidence that it is certainly related to a reduction in synaptic transmission [8,10, 8,10] and action potential desynchronization [37].
During exercise, the amplitude and frequency of hippocampal network activity increases [38, 39], along with neuronal firing [38] and synaptic transmission [23, 41]. However, this study is one of few showing that such functional changes can last longer than the exercise period [23, 43] and that they can even be observed ex vivo [23, 44]. It is known that exercise can induce long-lasting effects in several neuroprotective transduction pathways, which could contribute to the sustainably of exercise-induced changes in network function [21, 43] by producing a long-lasting protective state that would render the hippocampal network resistant to Aβ-induced disruptions [21, 40]. Despite the accruing evidence that exercise can have beneficial effects in AD patients and AD-models [40, 45], this is the first demonstration that it might do so by avoiding the deleterious effects of Aβ on network function. Thus, it is likely that exercise not only prevents Aβ production and accumulation [26], but it might prevent the action of Aβ itself on network function by inhibiting protein kinase GSK3β and by promoting PI3K activity.
Abundant evidence indicates that Aβ could activate intracellular signal cascades that ultimately activate GSK3β (for review see [11]). We, and others, have demonstrated that deleterious effects produced by Aβ involve the activation of GSK3β and that these disruptions can be reversed by the pharmacological inhibition of GSK3β with lithium or more specific inhibitors [12–14, 46]. These results agree with our finding that the specific inhibition of GSK3β with SB126763 prevents the inhibitory effect produced by Aβ on the CA1SPA of slices obtained from SED animals (Fig. 3), indicating that our original observation, that GSK3β is involved in the Aβ-induced inhibition of spontaneous population activity in the entorhinal cortex, is also true for the CA1 hippocampal subfield. The results also indicate that non-pharmacological manipulations that inhibit GSK3β can be successful against Aβ-induced deleterious effects and possibly against AD.
Interestingly, we also found that the pharmacological inhibition of PI3K eliminates the exercise-induced neuroprotective effect against Aβ-induced CA1SPA inhibition (Fig. 4). Abundant evidence also indicates that exercise leads to several neurochemical changes that converge in the activation of PI3K [36, 48] and the subsequent inhibition of GSK3β both in naïve [35, 49] and AD mouse models [16, 48]. In contrast to previous observations [50, 51], which show that PI3K inhibition by itself leads to GSK3β activation [50, 51], we found that the inhibition of PI3K in slices obtained from EX animals was not sufficient to activate GSK3β, but instead, it allows Aβ to activate this kinase and to inhibit CA1SPA (Fig. 4). We can only speculate that the neuroprotective effect of exercise is so strong that PI3K inhibition no longer results in GSK3β activation and that further stimulation, such as Aβ application, is required to activate GSK3β and to disrupt neural network activity after exercising. One possibility is that exercise changes GSK3β compartmentalization [52–55], as it does for glycogen-synthase [56], thereby modifying the regulation of GSK3β with respect to different transduction pathways or pathophysiological conditions [52–55]. In fact, it has been shown that changes in GSK3β compartmentalization interfere with PI3K-mediated cellular events [57].
In summary, we present evidence that voluntary exercise produces a long-lasting neuroprotective change that modifies hippocampal function and makes it resistant to deleterious insults, such as those induced by Aβ through GSK3β activation. The protective effect of exercise involves the reduction in GSK3β recruitment, possibly by the activation PI3K, which provides a mechanistic explanation for a promising, non-pharmacological approach to deal with AD as well as with other pathological conditions that involve disruptions in hippocampal network function.
Footnotes
ACKNOWLEDGMENTS
We thank Dr. Dorothy Pless for editorial comments, Dr. Benito Ordaz for technical assistance, and Dr. Gonzalo Martínez de la Escalera for providing access to his vivarium. Arturo González Isla “forma parte del programa de Doctorado en Ciencias Biomédicas, Universidad Nacional Autónoma de México (UNAM) y recibe actualmente la beca No. 514592 de CONACyT”. This study was supported by CONACyT Grants 117, 235789, 246888 and 181323; and by DGAPA-UNAM Grants IN200715 and IN201415.
