INSERM UMR1106, Institut de Neurosciences des Systèmes, Marseille, France
Aix-Marseille Université, Marseille, France
Address correspondence and reprint requests to Yuri Zilberter, Inserm UMR1106, Institut de Neurosciences des Systèmes, Faculté de Médecine La Timone, 27 Bd Jean Moulin, 13385 Marseille Cedex 05, France. E-mail: email@example.com
Deficient energy metabolism and network hyperactivity are the early symptoms of Alzheimer's disease (AD). In this study, we show that administration of exogenous oxidative energy substrates (OES) corrects neuronal energy supply deficiency that reduces the amyloid-beta-induced abnormal neuronal activity in vitro and the epileptic phenotype in AD model in vivo. In vitro, acute application of protofibrillar amyloid-β1–42 (Aβ1–42) induced aberrant network activity in wild-type hippocampal slices that was underlain by depolarization of both the neuronal resting membrane potential and GABA-mediated current reversal potential. Aβ1–42 also impaired synaptic function and long-term potentiation. These changes were paralleled by clear indications of impaired energy metabolism, as indicated by abnormal NAD(P)H signaling induced by network activity. However, when glucose was supplemented with OES pyruvate and 3-beta-hydroxybutyrate, Aβ1–42 failed to induce detrimental changes in any of the above parameters. We administered the same OES as chronic supplementation to a standard diet to APPswe/PS1dE9 transgenic mice displaying AD-related epilepsy phenotype. In the ex-vivo slices, we found neuronal subpopulations with significantly depolarized resting and GABA-mediated current reversal potentials, mirroring abnormalities we observed under acute Aβ1-42 application. Ex-vivo cortex of transgenic mice fed with standard diet displayed signs of impaired energy metabolism, such as abnormal NAD(P)H signaling and strongly reduced tolerance to hypoglycemia. Transgenic mice also possessed brain glycogen levels twofold lower than those of wild-type mice. However, none of the above neuronal and metabolic dysfunctions were observed in transgenic mice fed with the OES-enriched diet. In vivo, dietary OES supplementation abated neuronal hyperexcitability, as the frequency of both epileptiform discharges and spikes was strongly decreased in the APPswe/PS1dE9 mice placed on the diet. Altogether, our results suggest that early AD-related neuronal malfunctions underlying hyperexcitability and energy metabolism deficiency can be prevented by dietary supplementation with native energy substrates.
The amyloid cascade hypothesis dominates the AD research. It posits that accumulation of amyloid-β peptide (Aβ) in the brain results in neuronal dysfunction, synaptic loss, and cognitive decline in AD (Karran et al. 2011; Selkoe 2011). Therefore, most research efforts target processes of either Aβ production (Small and Duff 2008; Hardy 2009) or clearance (Kurz and Perneczky 2011; Miners et al. 2011). However, targeting Aβ for clinical treatment has so far failed to show promising results (Saxena 2010; Karran et al. 2011; Huang and Mucke 2012). One obvious reason for the failure is that all trials simply started too late in the disease course. According to the present view, derived mainly from the advent of positron emission tomography (PET) imaging with an amyloid-binding ligand, amyloid pathology may start up to two decades before the symptomatic phase of the disease (Ashe and Zahs 2010) and there is little amyloid accumulation in the brain in diagnosed AD patients (Engler et al. 2006). This suggests that molecular pathways downstream of Aβ accumulation may provide better treatment targets than Aβ itself. Hypometabolism, which manifests as reduced glucose metabolism, neuronal energy deficit, and oxidative stress (Mosconi et al. 2009; Cunnane et al. 2008; Murray et al. 2011) may represent such a pathway. Importantly, while these phenomena are observed early on in the pathogenesis of AD (Frackowiak et al. 1981), they also accompany the disease progression (Engler et al. 2006; Saxena 2011). Research into metabolic abnormalities may thus translate into treatment options throughout the course of AD.
Several major molecular abnormalities in AD are directly associated with the state of energy metabolism (Velliquette et al. 2000; Guglielmotto et al. 2009b; Avila et al. 2012; Vassar and Kandalepas 2005). For instance, inhibition of neuronal energy supply leads to the elevation in BACE1 levels and its activity and subsequent Aβ overproduction. In addition, chronic stress and hypoxia increase both β-secretase and γ-secretase activity, also resulting in Aβ overproduction (Guglielmotto et al. 2009b; Vassar and Kandalepas 2005). Furthermore, increased activity of a key enzyme in cellular glycogen metabolisms, glycogen synthase kinase 3, results in both tau hyperphosphorylation and increased Aβ generation (Avila et al. 2012). In turn, Aβ can induce oxidative stress (Manczak et al. 2006; Abeti et al. 2011) by sequential activation of NADPH oxidase in glia, reactive oxygen species generation, astrocyte-specific cytosolic Ca2+ overload, up-regulation of activity of DNA repair protein poly(ADP-ribose) polymerase (PARP)-1, sequestration of critical metals important for metabolism, and depletion of cytosolic NAD+ (Abeti et al. 2011). This results in the inhibition of glycolysis, depletion of mitochondrial respiratory substrates, energy deficiency, and neuronal damage (de la Torre 2006; Cunnane et al. 2008; Kapogiannis and Mattson 2011). Aβ can also directly impair neuronal glucose uptake (Mark et al. 1997; Prapong et al. 2002), glycolysis (Bigl and Eschrich 2001), and mitochondrial function (Casley et al. 2002; Hong et al. 2007; Tillement et al. 2011). Altogether, a large body of evidence indicates a strong link between impaired energy metabolism and early pathogenesis in AD.
As most neuronal processes, including those underlying electrophysiological properties, are energy dependent, disruptions in energy metabolism should manifest in altered neuronal activity. Recent studies in mice and rats point to a causal link between energy metabolism and neuronal excitability (Wang et al. 2012b, b). However, this area remains largely unexplored, especially in the field of AD research. Hippocampal hyperactivity has been reported in human subjects at high risk for developing AD, such as pre-symptomatic carriers of familiar AD-linked genes (Quiroz et al. 2012) or APOE-epsilon4 allele (Filippini et al. 2009), or subjects with mild cognitive impairment (Dickerson et al. 2005; Filippini et al. 2009; Bakker et al. 1995; Quiroz et al. 2012). Epidemiological studies also point to the increased risk of unprovoked seizures in AD patients compared with age-matched population (Amatniek et al. 2010; Noebels 2011). Recent studies, including our own, have revealed that epileptic seizures are frequent in transgenic mice that develop amyloid plaque pathology (Palop et al. 2009; Minkeviciene et al. 2009; Palop and Mucke 2007). Further in vitro studies have shown that the abnormal presence of Aβ is responsible for neuronal hyperexcitability (Palop et al. 2009; Minkeviciene et al. 2009; Cuevas et al. 2011) that may account for the observed unprovoked seizures in AD.
As metabolic dysfunction and neuronal hyperexcitability are linked (Pan et al. 2008; Kudin et al. 2009; Waldbaum and Patel 2012a), we hypothesize that exogenous energy substrate compensation can overcome cellular energy deficiency and prevent disruption of neuronal function in AD, particularly neuronal excitability. Provided that Aβ inhibits neuronal glucose transport (Mark et al. 1997; Prapong et al. 2002), bypassing the glucose-dependent pathways by administration of oxidative energy substrates (OES) for mitochondrial respiration would abate the toxic effects of Aβ on neuronal energy metabolism. To verify our hypothesis, we used either energy substrate-enriched artificial cerebro-spinal fluid (eACSF: ACSF supplemented with 5 mM pyruvate and 4 mM DL-3-beta-hydroxybutyrate (Holmgren et al. 2010)) for the in vitro experiments or energy substrate-enriched diet (standard diet supplemented with pyruvate and 3-beta-hydroxybutyrate) for the in vivo experiments. We employed a widely used transgenic AD mouse model over-expressing mutated human amyloid precursor protein and presenilin-1 (APPswe/PS1dE9). In these mice, Aβ levels, especially those of Aβ1–42, are elevated already at a young age and first amyloid plaques are observed in the cortex and hippocampus around 4 months of age (Garcia-Alloza et al. 2006), while signs of hyperexcitability and seizures are detected after 3 months of age (Minkeviciene et al. 2009).
Materials and methods
In vitro experiments
Tissue slice preparation
Brain slices were prepared from Swiss mice of both sexes of different ages (from 3 to 25 weeks). All animal protocols conformed to the French Public Health Service policy and the INSERM guidelines on the use of laboratory animals. The mouse was killed, rapidly decapitated and the brain was removed from the skull and placed in the ice-cold ACSF oxygenated with 95% O2/5% CO2. The ACSF solution consisted of (in mmol/L): NaCl 124, KCl 2.50, NaH2PO4 1.25, NaHCO3 25, CaCl2 2.00, MgSO4 1.30, and dextrose 10, pH 7.4. Saggital slices (350 μm) were cut using a tissue slicer (Leica VT 1200s; Leica Microsystem Vertrieb GmbH, Wetzlar, Germany). During cutting slices were submerged in an ice-cold (< 6°C) cutting solution consisting of (in mmol/L): K-gluconate 140, HEPES 10, Na-gluconate 15, EGTA 0.2, NaCl 4, pH adjusted to 7.2 with KOH. Slices were transferred immediately to an oxygenated holding chamber maintained at 22°C, and allowed to recover for 2 h. Slices were then transferred to a standard round 1.5-mL recording chamber and submerged (~2 mm) in ACSF buffer, which was continuously superfused (10–15 mL/min) and oxygenated with 95% O2/5% CO2. The temperature in the chamber was kept at 33–34°C for all experimental conditions.
Acute amyloid-β application
Protofibrillar Aβ 1–42 was synthesized and prepared as described previously (Minkeviciene et al. 2009). Before use, the solid, fibrillar peptide was suspended in ACSF or eACSF, sonicated for 1 min, and applied within 60 min at a final concentration of 1.0 μM.
Synaptic stimulation and field potential recordings
Shaffer collateral/commissural pathway was stimulated using the DS2A isolated stimulator (Digitimer Ltd, Welwyn Garden City, UK) with a bipolar nichrome electrode situated in the stratum radiatum of CA1 hippocampal region. Stimulus current was adjusted using single pulses (170–240 μA, 200 μs, 0.15 Hz) to produce a local field potential (LFP) of nearly 50% of maximal amplitude. LFPs were recorded using glass microelectrodes filled with ASCF, placed in stratum pyramidale, and connected to the DAM-80 amplifier (WPI, Sarasota, FL, USA). An extended synaptic stimulation consisted of a 10-s or 20-s stimulus train (200-μs pulses at 10 Hz) was used to generate autofluorescence reduced pyridine nucleotide response. Long-term synaptic plasticity (LTP) was induced by two 100-Hz, 1-s stimulations (separated by 1-min interval) of Schaffer collaterals and measured as a slope of field potentials recorded in stratum radiatum. The stimulation intensity was chosen to be at least 50% smaller than that inducing a population spike.
Gramicidin patch recordings
Gramicidin (Sigma, St. Louis, MO, USA) was dissolved in dimethylsulfoxide (DMSO; 50 mg/mL). The antibiotic–DMSO solution was added to the electrode solution to a final concentration of 50–100 μg/mL. After seal formation, the progress of perforation was monitored by the current transient in response to a 5-mV step in holding potential. The recordings were initiated once the input resistance stabilized (200–400 MΩ). GABA currents were induced by a brief (10-ms) puff of 100 μM GABA-containing ACSF onto the somatic region of the neuron at a range of membrane holding potentials. Both peak GABA current and membrane current amplitudes were measured and plotted on the IV curve with a linear fit to determine their respective reversal potentials.
NAD(P)H fluorescence imaging
Reduced nicotinamide adenine dinucleotide phosphate (NADPH) and reduced nicotinamide adenine dinucleotide (NAD(P)H) have very similar optical properties, and therefore is expected that NADPH also contributes to some extent to total autofluorescence signals (Klaidman et al. 1995; Shuttleworth 2010). As the two components are recorded as a single signal, this signal is referred to as NAD(P)H. Changes in NAD(P)H fluorescence in hippocampal slices were monitored using a 290- to 370-nm excitation filter and a 420-nm-long pass filter for the emission (Omega Optical, Brattleboro, VT, USA). The light source was the Intensilight C-HGFI illuminator (Nikon Instruments Europe B.V., Amstelveen, The Netherlands) equipped with a mercury arc lamp. Slices were epi-illuminated and imaged through a Nikon upright microscope (FN1, Eclipse) with 4x/0.10 Nikon Plan objective. Images were acquired using a linear, cooled 12-bit CCD camera (Sensicam, PCO AG, Germany) with a 640 × 480 digital spatial resolution. Because of a low level of fluorescence emission for this fluorophore, NAD(P)H images were acquired every 500–600 ms as 8 × 8 binned images (effective spatial resolution of 80 × 60 pixels). The exposure time was adjusted to obtain fluorescence intensity between 2000 and 3000 optical intensity levels. The images were stored on a computer as 12-bit files (0–4096 dynamic range). Fluorescence intensity changes in stratum radiatum near sites of LFP and O2 recordings were measured in three to five regions of interest using ImageJ software (developed by Wayne Rasband, NIH, Bethesda, Maryland, USA). Data were expressed as the percentage changes in fluorescence over a baseline [(ΔF/F) 100]. Signal analysis was performed using IgorPro software (WaveMetrics, Inc., Lake Oswego, OR, USA).
A Clark-style oxygen microelectrode (OX-10, tip diameter 10 μm; Unisense Ltd, Århus, Denmark) was used to measure slice tissue PO2. The electrode was connected to a picoammeter (PA2000, Unisense Ltd) and the cathode was polarized at 800 mV in normal saline at 22°C for up to 12 h before the first use. A two-point calibration (in pA) was performed following polarization by inserting the electrode in normal saline solution (at 33°C) equilibrated with either 95% O2 5% CO2 or ambient air. Calibrations were repeated after each experiment to determine the PO2 values. The oxygen electrode was positioned using motorized micromanipulator (Scientifica Ltd, Uckfield, UK) in the proximity to the field potential recording electrode.
Drugs used were purchased from Sigma, St. Louis, MO, USA (racemic mixture of DL-3-hydroxybutyric acid sodium salt, pyruvate sodium salt). Within the racemic mixture, D-BHB is the primary mediator of the physiological effects of DL-BHB, and is the only form that can function as a substrate for mitochondrial BHB dehydrogenase. Consequently, only 50% of exogenous DL-BHB is expected to be utilized (Tsai et al. 2011).
Group measures were expressed as means ± SEM; error bars also indicate SEM. Statistical significance was assessed using the Wilcoxon's rank sum test or Student' paired t-test (unless otherwise noted). The level of significance was set at p <0.05. Graphical representation of significance: *** denotes < 0.001 significance; ** denotes < 0.01 significance; and * denotes < 0.05 significance.
Calculating LFP integrals
The procedure of calculating LFP integrals in the train has been previously described in detail (Ivanov et al. 2011). During analysis of the LFP train, the computer program separated each LFP, shifted the baseline to 0, and selected the region of integration. Population spikes were inverted and then the integral of the whole trace was calculated.
In vivo experiments
Female transgenic APPswe/PS1dE9 mice (Jankowsky et al. 2004) were used in the study. The mice came from a local colony at the University of Eastern Finland in Kuopio, based on breeders from Johns Hopkins University (Baltimore, MD, USA). The mice carried mouse/human APPswe double point mutations and human presenilin-1 gene with deleted exon 9, cointegrated in the same transgene under the mouse PrP promoter. This line was originally maintained in a hybrid C3HeJ x C57BL6/J F1 background, but the mice used in this study were derived from backcrossing to C57BL6/J for 14 generations. Mice were housed under 12:12-h light/dark cycles, with food and water available ad libitum. Experiments were conducted in accordance with the European Communities Directive (86/609/EEC), and approved by the National Animal Experiment Board, State Provincial Office of Southern Finland.
APPswe/PS1dE9 mice (n =9, age 12–13 week) were treated with pyruvate and 3-β-hydroxybutyrate (BHB)-enriched chow for 5 weeks (RM1 + 0.33% NaPyruvate + 0.33% BHB; Special Diets Services, Nova SCB AB, Sollentuna, Sweden). Thus, the average daily consumption of substrates was about 26 mg. Before and after the dietary intervention, mice were fed with regular chow (RM1). The control group (n =8) received regular food/chow (RM1) throughout the experiment. Pyruvate – BHB-enriched diets were available ad libitum and replaced twice a week.
For video-EEG monitoring, cortical screw electrodes were implanted under general anesthesia (isoflurane, 1.5–2.5%) as described previously (Minkeviciene et al. 2009). Mice were allowed to recover for 12 ± 2 days before the baseline recording was started.
Video-EEG monitoring and analysis
To evaluate the effect of the diet with alternative energy sources on spontaneous epileptiform activity, continuous (24/7) video-EEG recording was performed for 1 week before the diet, during the last 2 weeks of a 5-week dietary intervention, and for 2 weeks after returning to regular food. Video-EEG monitoring was performed as previously described (Minkeviciene et al. 2009). Mice were housed individually in Plexiglas cages where they could move freely and connected to a Nervus EEG recording system (sampling rate 256 Hz, high-pass filter 0.5 Hz, low-pass filter 100 Hz). EEG analysis was done visually by scanning through each EEG file on the computer screen. As only two mice in the diet group and one control mouse had spontaneous seizures at baseline, epileptiform discharges (ED) were counted for 1-week epochs for all video-EEG data obtained. An ED was defined as a high-amplitude (> 2x baseline) rhythmic event with duration between 1 and 5 s. (Fig. 2bii). In addition, epileptiform spikes (ES) were counted for 4-h epochs (20:00–24:00) on days 5, 6, and 7 of baseline EEG recording, on the same days of the 5th week of the dietary intervention, and of the 2nd week after the diet. An ES was defined as a high-amplitude (twice the baseline) sharply contoured waveform with a duration of 20–70 ms. (Fig. 2biii).
Brain glycogen concentration measurements
Whole brain hemispheres were extracted from WT and TG mice and snap frozen in liquid nitrogen, then crushed and placed in acidified ethanol overnight. Samples were then filtered out, placed into a solution containing 0.1 M NaOH, 0.01% sodium dodecyl sulfate , and 1 mM EDTA, homogenized to form uniform suspension and neutralized to pH 7 with HCl. Supernatant was removed after centrifuging at 5000 rpm for 15 min and divided into three sample groups: a) for protein assay, b) for free (not dissolved by ethanol) glucose measurements, and c) for glycogen concentration measurements. Glycogen was hydrolyzed in (c) group, using Amyloglucosidase (Sigma) as a catalyst. Glucose content was then measured in groups (b) and (c) using the HK Glucose assay kit (Sigma) as per manufacturer instructions. Initial sample glycogen content was then calculated using the glucose measurements, considering 1–1 mol conversion of glycogen into glucose in our samples (with free glucose subtraction) and using a conversion factor of 181 g/mol free glucose = 162 g/mol glucosyl units of glycogen (Abdelmalik et al. 2007). Glycogen content was normalized to sample protein levels as analyzed in group (a) using Precision Red protein assay kit (Cytoskeleton, Denver, CO, USA) as per manufacturer's instructions. The glycogen units are presented as μmol glycosyl unit/g wet tissue weight by considering the fact that protein content in the brain constitutes 11.7% of wet brain weight (Choi and Gruetter 2003).
ELISA assays for amyloid-β levels
Brain tissue samples from the parieto-occipital cortex were weighed and homogenized in 8x the original volume of 5 M guanidine-HCl/50 mM Tris·HCl, pH 8.0 and mixed on a shaker for 3 h at 22°C. The sample was further diluted in 20x volume of Dulbecco's phosphate-buffered saline buffer (Sigma) with 5% bovine serum albumin , containing complete inhibitory mixture (Roche Diagnostics, Basel, Switzerland, Germany) and centrifuged at 16 000 g for 20 min at 4°C. Decanted supernatant was further diluted at 1 : 200 with dilution buffer. Diluted samples were then used to analyze total Aβx-42 species. Aβ42 levels were estimated using ELISA kits (Biosource International, Grand Island, NY, USA) in accordance with manufacturer's instructions. The levels were standardized to brain tissue weight and expressed as pg of Aβ per mg ± SEM.
Within each experimental group, the effect of diet on the number of EDs and ES was assessed with Wilcoxon signed-rank test. The intergroup differences at baseline and on diet were assessed by Chi-Square test (number of responders and non-responders). Responders were defined as animals that had at least a 50% reduction in number of EDs/spikes during/after the diet as compare with baseline. Statistical analyses were performed using SPSS 14.0 (IBM software, New York, USA) and Excel.
Energy substrate supplementation prevents Aβ-induced aberrant network activity, modulation of synaptic function and NAD(P)H signaling
Although Aβ has been implicated in AD-related epileptogenesis (Roberson et al. 2011), it was not evident whether Aβ itself could induce paroxysmal activity in vitro. Therefore, we recorded field network activity in the area CA1 of acute hippocampal slices of wild-type mice before and during a 1-h protofibrillar Aβ1–42 exposure. As we have previously reported, fibrillar Aβ, but not oligomers induced neuronal hyperactivity (Minkeviciene et al. 2009; see also Busche et al. 2012). In most experiments (12 of 14), interictal-like spontaneous synchronized discharges appeared following 40 min of Aβ1–42 application (Fig. 1a). These discharges were not observed in Aβ-free ACSF (Fig. 1a). Interestingly, in the presence of eACSF, none of the slices displayed synchronized discharges in response to Aβ1–42 application (Fig. 1b; n =5). These data suggest that protofibrillar Aβ induces aberrant network activity, which is prevented by OES supplementation.
As Aβ-induced network hyperactivity was abated by OES, we investigated whether it is associated with Aβ modulation of energy metabolism. We induced neuronal activity in the CA1 region by Schaffer collateral stimulation (10 s, 10 Hz) in wild-type hippocampal slices, and recorded the resulting NAD(P)H autofluorescence and partial oxygen pressure (pO2) changes before and after Aβ1–42 application. NAD(P)H transients displayed a typical biphasic shape characterized by a dip (oxidation phase) followed by an overshoot (Fig. 2ai). The initial NAD(P)H decrease represents enhanced NAD(P)H oxidation in mitochondria, while a substantial part of the overshoot likely results from the NAD(P)H generation during glycolysis [(Kasischke et al. 2004); A.I. & Y.Z., unpublished] followed by a decline to basal level in parallel with a decrease in energy demands, dehydrogenase activity, and lactate uptake (Shetty et al. 2012a). Following 1-h Aβ1–42 application, NAD(P)H signal was strongly modified (Fig. 2ai): Aβ1–42 increased the oxidation area (p <0.02) without affecting the oxidation phase amplitude (p >0.1), and reduced both the overshoot area and amplitude (Table 1i, j; p <0.001). Meanwhile, the total oxygen consumption did not change significantly (Table 1h; p >0.3). Therefore, Aβ1–42 effect on network excitability was paralleled by significant alterations in metabolic parameters.
Table 1. OES supplementation prevents the acute effects of Aβ on neuronal energy metabolism and excitability parameters in slices from WT mice
ACSF + Aβ
eACSF + Aβ
Total LFP train integral is the summation of all the LFPs in a single 10-Hz, 10-s stimulus train.
The stimulation protocol we utilized for NAD(P)H imaging uncovered yet another Aβ effect on synaptic function. Following Aβ1–42 application, local field potentials (LFPs) within the stimulation train decayed faster than those recorded in the absence of Aβ (Fig. 2aii). To quantify changes in LFP dynamics, we used total train LFP integrals (see Methods), because as a result of short-term synaptic plasticity, LFPs are strongly modulated in the train (Pitler and Landfield 1987) (Fig. 2aii) and therefore cannot be informatively characterized by their amplitude. LFP integrals in the train confirmed the faster decay dynamics for Aβ1–42-treated slices (Fig. 2aiii; n =10). This difference was also revealed by the total LFP integral (summation of all LFP integrals in the train): after Aβ1-42 application, it was reduced by 14% (p <0.01; Table 1g). These findings imply that in addition to disrupting energy metabolism parameters, Aβ1–42 also impairs the capacity of synapses to maintain their function during prolonged periods of activity.
In addition, we tested the effect of Aβ on LTP as this issue is of primary importance in pathogenesis of AD. In agreement with previously reported results (Koffie et al. 2011), Aβ1–42 induced a strong decrease in the LTP level (n =10) compared with the untreated slices (p <0.01); n =9; Fig. 3).
Next, we performed similar protocols to see whether eACSF can reverse Aβ-induced impairment in neuronal energy metabolism. Contrary to the results obtained in regular ACSF, we now observed no Aβ-induced changes in metabolic parameters (Fig. 2bi): Aβ1–42 caused no significant changes either in the oxidation area (p >0.1) or amplitude (p >0.5), or in the overshoot area (p >0.3) or amplitude (p >0.5). The oxygen consumption did not change either (Table 1h–j; p >0.7). In addition, Aβ affected neither recorded LFP dynamics (Fig. 2bii,c; n =5) nor total LFP integrals (Table 1g; p >0.9). Moreover, Aβ1–42 treatment in the presence of OES did not induce any changes in LTP (n =7) compared with the untreated slices (p >0.3; Fig. 3). Therefore, supplementing glucose with OES prevents the toxic action of Aβ on both the energy metabolism and synaptic function.
OES prevent the toxic effect of Aβ on neuronal excitability
Aberrant network activity in the presence of Aβ may be associated with modifications of basic neuronal parameters underlying neuronal excitability. Using gramicidin patch recordings that leave the intracellular chloride intact, we measured in slices the reversal potential of GABA-induced currents (EGABA) and the resting membrane potential (Em), recording these parameters in the same DG granule neuron before and after 1-h exposure to protofibrillar Aβ1–42. In WT slices superfused with ACSF, consistent with our previously reported results (Minkeviciene et al. 2009), Aβ1–42 significantly depolarized the Em (p <0.0005; Fig. 4aii, iii; Table 1a), but also EGABA (p <0.005; Fig. 4ai, ii, iv; Table 1b). We confirmed similar effects of Aβ1–42 on pyramidal cells in neocortical layer 2/3 using a non-invasive technique of cell-attached NMDA and GABA single-channel recordings (Rheims et al. 2009) (Em, p <0.0001, Fig. 4bi, Table 1c; and EGABA, p <0.0004; Fig. 4bii, Table 1d). Finally, Aβ-induced hyperexcitability was also verified as the change in CA1 population spike (PS) integral in response to a single-pulse stimulation of Schaeffer collaterals. In ACSF, the PS integral increased twofold following Aβ1–42 treatment in slices of both 3-week-old (p <0.004; Fig. 4c, Table 1e) and 3-month-old WT mice (p <0.02; Fig. 4c, Table 1f). Altogether, these data indicate that Aβ may induce hyperexcitability in various cell types and different brain areas via at least two excitability-underlying neuronal parameters: by depolarizing Em, Aβ facilitates the generation of action potentials, and by depolarizing EGABA, Aβ weakens the inhibitory effect of GABAergic inputs.
To confirm that OES supplementation may prevent Aβ-induced neuronal excitability, we repeated our experiments in the presence of OES. Indeed, in eACSF, Aβ1–42 affected neither Em nor EGABA in both DG granule cells (p >0.05 for each; Fig. 4aiii, iv) and L2/3 pyramidal cells (p >0.09 and p >0.9, respectively; Fig. 4c, Table 1a–d). Finally, Aβ1–42 failed to induce any significant changes in the CA1 population spike (p >0.3 and p >0.6; Fig. 4d, Table 1e, f). Therefore, we find that acute OES supplementation prevents the toxic action of Aβ on neuronal excitability parameters.
OES supplementation improves brain energy metabolism in APP/PS1 mice
Our in vitro results suggested that Aβ can cause neuronal hyperexcitability and energy deficit, and that these effects are abated by OES in acute brain slice application. However, it remained to be seen whether these results are valid in the in vivo mouse model of AD. To this end, we administered a combination of pyruvate and BHB to mice as a chronic supplementation to a standard high-carbohydrate diet. TG mice (~3-months old) were placed for 5 weeks on a diet supplemented with OES (TGOES group) or a non-supplemented standard diet (TGSTD group). Two groups of wild-type littermates (WTSTD and WTOES) were placed on similar diets as well. Thereafter, we compared parameters of neuronal activity in the ex vivo slices perfused with regular glucose-containing ACSF.
First, to test whether TG mice also display impairment of energy metabolism, we prepared hippocampal slices from TG and WT mice. We induced neuronal activity in the CA1 region by Schaffer collateral stimulation (30 s, 10 Hz) and recorded the resulting NAD(P)H autofluorescence. In most cases, the wave shape of NAD(P)H transients in TGSTD slices noticeably differed from those in WT mice, especially in a smaller overshoot/oxidative phase ratio (Fig. 5a, Table 2c). Interestingly, similar pathological changes in NAD(P)H autofluorescence have been reported in tissue from patients with temporal lobe epilepsy exhibiting pronounced neuronal hypometabolism (Kann et al. 2005). In TGOES mice, NAD(P)H transients did not differ significantly from WT mice (Fig. 5a, Table 2c). Therefore, NAD(P)H imaging uncovered apparent abnormal energy metabolism in TGSTD mice, a pathological phenomenon not observed in TGOES mice.
Table 2. OES diet normalizes neuronal energy metabolism and excitability parameters in AD mouse model
TG standard diet
TG OES diet
EGABA, Dentate Gyrus granule cells
−80.2 ± 1.0 mV
−73.9 ± 1.1 mV
−77.3 ± 0.8 mV
Em, Dentate Gyrus granule cells
−88.2 ± 1.0 mV
−81.0 ± 5.0 mV
−90.3 ± 0.5 mV
NAD(P)H Oxidation to Overshoot Amplitude Ratio
2.81 ± 0.38
1.96 ± 0.25
3.33 ± 0.51
Time to 50% PS decay in 0.1 mM glucose
34 ± 2.8 min
22.3 ± 3 min
31.8 ± 2.8 min
Brain glycosyl units content, μmol per g wet weight
5.54 ± 1.03
2.52 ± 0.79
4.76 ± 2.04
Presumably as a consequence of disturbed energy metabolism, neurons in hippocampal slices of TGSTD mice showed significantly lowered resistance to glucose deprivation. We employed an experimental protocol similar to that previously used for testing glycogen support of neuronal activity during energy deprivation (Brown and Ransom 2012). In short, the protocol consisted of recording CA1 PS integral in normal 10 mM glucose-containing ACSF and counting the time to its decay to 50% after switching to 0.1 mM glucose-containing ACSF. To intensify energy metabolism, high network activity was induced by an additional 20-pulse, 100-Hz stimulus train delivered once a minute. In WT mice, OES-enriched diet had no effect (34 min for both WTSTD and WTOES groups, p >0.5; Fig. 5b, Table 2d). However, the OES diet produced a pronounced effect in TG mice, which under the standard diet exhibited significantly lower resistance: the PS decayed to 50% within 22 min in TGSTD mice, while it took 31 min in TGOES mice (p <0.05; Fig. 5b, Table 2d). Thus, the neuron synaptic function in TGSTD displayed a strongly reduced tolerance to energy deprivation, a pathological signature that was significantly abated by the OES diet.
One possible reason for the fast collapse of synaptic function under extracellular glucose deficiency may be in a reduced cellular ability to generate the intracellular energy buffers. Normally, glycogen represents the endogenous energy store in the brain and is predominantly located in astrocytes (Brown and Ransom 2012). Mobilization of energy from glycogen can be triggered by neuronal activity and the consequent release of glutamate (Brown and Ransom 2012; Obel et al. 2008). In AD, glycogen buffers can be depleted because of limited glucose utilization and hyperactivity of GSK-3 (Avila et al. 2012). We therefore measured the glycogen content in the brain homogenates of mice on OES versus STD chow. The measurements showed a strong reduction of glycogen levels in TGSTD mice compared with WT mice (on either OES or STD diet). However, such reduction was not observed in TGOES mice (Fig. 5c, right; Table 2e). These results suggest that dietary supplementation with OES improves glycogen replenishment and prevents depletion of glycogen stores of TG mice, thus increasing their ability to sustain high network activity (Brown and Ransom 2012; Sickmann et al. 2009; Shetty et al. 2012b).
As we observed in case with acute Aβ application in WT slices, disturbed energy metabolism in TGSTD mice likely affects basic neuronal parameters underlying excitability, which may be one reason for aberrant neuronal activity described in these mice in vivo (Minkeviciene et al. 2009). In slices from TGSTD mice, we also recorded, although infrequently (four from 12 slices, five mice), interictal-like spontaneous discharges (Fig. 5d) resembling those seen in WT slices after Aβ application (Fig. 1a). These discharges were not observed in WT and TGOES mice.
We also measured neuronal Em and EGABA in ex vivo slices from all mouse groups. DG granule cells from TGSTD mice exhibited a significantly depolarized EGABA (p <0.001, Table 2a), another abnormality similar to that observed after acute Aβ application. Moreover, the EGABA distribution in these TGSTD cells was fitted well by two Gaussians (Fig. 5e) indicating the presence of two neuronal subpopulations: one with EGABA similar to that in WT mice and another one (about 40% of cells) where EGABA was depolarized by about 10 mV. TGOES mice did not exhibit such abnormalities (Fig. 5e, Table 2a). In addition, Em in TGSTD mice was depolarized by ~7 mV compared to WT mice (Fig. 5f; p <0.05, Table 2b), a phenomenon that was not observed in TGOES mice (p >0.3, Table 2b). Meanwhile, the OES diet induced no effect either on EGABA (p >0.1) or Em (p >0.5) in WT DG granule cells. TGSTD mice thus appear to possess a subpopulation of hyperexcitable cells because of their depolarized Em or/and EGABA, factors that may contribute to network hyperexcitability and epilepsy (Cohen et al. 2011). Critically, in TG mice placed on the OES-enriched diet, these parameters exhibited values similar to those observed in WT mice.
OES supplementation decreases epileptiform activity in APP/PS1 mice
The effect of OES-supplemented diet on network hyperexcitability was paralleled by the reduced epileptiform activity in vivo. We placed TG mice (age 12–13 weeks) on the OES-enriched (n =9) or control (n =8) diets for 5 weeks (Fig. 6a). TG mice displayed epileptiform EEG activity in the form of spontaneous seizures (Fig. 6b), epileptiform discharges (EDs) (Fig. 5c), and epileptiform spikes (ES) (Fig. 6d). During the recorded 4th and 5th weeks on OES-enriched diet, the mean number of EDs in TGOES mice was reduced by 67 ± 31% (mean ± SD) as compared with baseline (p <0.05, Fig. 6c), while TGSTD mice showed no change in EDs (p >0.05; Fig. 6c). Interestingly, the effect was long lasting, as the EDs remained reduced even throughout the 2-week post-diet period in the TGOES group (p <0.05), but not in TGSTD controls (Fig. 6c). In particular, the prolonged effect was seen in mice with low frequency of EDs at baseline (median ≤ 7 EDs/week) as compared with those with more frequent EDs (> 7, p <0.05, chi-square). OES-enriched diet also reduced the ES frequency by 59 ± 25% as compared with baseline (p <0.05, Fig. 6d). Unlike the case with EDs, the dietary effect on ES disappeared following the discontinuation (p >0.05, Wilcoxon signed-rank test; Fig. 6d). However, during the 2-week period after discontinuation of the diet, 43% of the TGOES mice showed a reduced frequency of spikes, while none of the TGSTD mice displayed any substantial decrease in spike frequency (p <0.05, chi-square; Figure S1f).
Importantly, as was shown by the ELISA test, the Aβ1–42 levels in the TGSTD (n =7) and TGOES (n =9) groups did not differ significantly (p >0.9), being at 1348 + 122 and 1355 + 104 pg/mg, respectively. This suggests that the effect of OES-supplemented diet is most likely underlain by the prevention of toxic Ab action, rather than by the reduction of Ab levels.
Overall, the OES-supplemented diet resulted in a marked decrease in in vivo epileptiform activity of TG mice that was paralleled by normalized basic parameters underlying neuronal excitability and energy metabolism.
In this study, we show that correction of impaired brain energy metabolism can abate one of the earliest signs of AD-related neuronal dysfunctions, hyperexcitability. Our results indicate that Aβ can acutely affect basic parameters of brain energy metabolism, and presumably through this pathway induces fundamental changes in electrophysiological properties of neurons, such as those underlying network excitability. Compensation of neuronal energy supply deficiency by exogenous OES largely eliminated the toxic action of Aβ on neuronal excitability in vitro and strongly reduced network hyperactivity of TG mice in vivo.
The link between Alzheimer's disease and brain energy metabolism has been known for decades, as disruption of brain glucose utilization in PET imaging is the earliest surrogate marker of AD (Frackowiak et al. 1981). However, the causality between energy hypometabolism and progression of AD remains unclear (Blass 2007; Cunnane et al. 2008; Saxena 2011). This relationship is complicated by the fact that deficient energy metabolism favors production of Aβ (Velliquette et al. 2000; O'Connor et al. 2012; de la Torre 2006; Guglielmotto et al. 2009a), which further impairs the energy supply, thus creating a vicious cycle. On the other hand, cardiovascular disease or its risk factors predispose to AD according to epidemiological studies (Stampfer 2006; Patterson et al. 2007; de la Torre 2006), and impaired glucose metabolism can be detected in conditions with high-risk for conversion to AD, such as presence of APOE-epsilon4 allele, family history of AD, or diagnosed mild cognitive impairment (Mosconi et al. 2009; Cunnane et al. 2008). It has also been proposed that brain energy deficiency may be a primary player in the disease initiation (Blass 2007; Cunnane et al. 2008; Saxena 2011). In support to this notion, impaired mitochondrial function can be detected in AD model mice before any amyloid deposition into plaques (Hauptmann et al. 2009; Yao et al. 2010).
Neuronal hyperactivity leading to abnormal oscillations and epilepsy has been observed in different mouse models of AD (Palop and Mucke 2007). One possible underlying mechanism for such paroxysmal network activity may be deficits in parvalbumin-positive inhibitory interneurons according to a recent report (Verret et al. 2010). We found, in line with our previous results (Minkeviciene et al. 2009), that the general reason for hyperactivity may be the Aβ-induced modification of basic neuronal properties, such as Em and EGABA. Indeed, depolarization of both parameters in excitatory cells was associated with synchronized spontaneous network activity in brain slices (see Figs 1, 4). Similar events were also observed in the ex vivo slices from TG mice (see Fig. 5). These abnormalities were paralleled by clear indications of energy metabolism imbalance. Critically, compensation for energy deficiency by acute OES application (in vitro) or OES dietary supplementation (in vivo) largely prevented these abnormalities. In vivo, this was also evidenced by a considerably reduced epilepsy phenotype in TGOES mice (Fig. 6s).
In vivo, the normalization of energy metabolism was indicated by a twofold higher level of brain glycogen in the TG mice fed with OES-enriched diet compared with the control TG mice
Indeed, the content of glycogen in TGOES mice did not differ from that in WT mice, while it was considerably lower in TGSTD mice. The role of glycogen as the primary energy buffer in the brain has been underestimated for a long time. Recent studies revealed, however, that glycogen can be present in large amounts in the brain of both rodents and humans (Oz et al. 2009, 2007) where it serves as an endogenous energy source stored in astrocytes. Physiological activation increases astrocytic glycogen utilization during normal brain activation, as well as when oxygen or glucose supply is inadequate (Dienel and Cruz 2006). Moreover, brain glycogen supports energy metabolism when glucose supply from the blood is inadequate and initially reduced glycogen levels rebound to levels higher than normal after a single episode of moderate hypoglycemia in humans (Oz et al. 2007). According to the astrocyte–neuron lactate shuttle model (Pellerin and Magistretti 2011), lactate derived from glycolysis shuttles from astrocytes into neurons where lactate dehydrogenase converts it back to pyruvate to be used in oxidative respiration in mitochondria. The compromised astrocytic glycogen stores in TG mice can result from impaired glucose transport and uptake (Mark et al. 1997; Prapong et al. 2002), impaired glycolysis (Hoyer 2004), or hyperactivated GSK-3 (Cohen 2002; Avila et al. 2012). One can speculate that the use of pyruvate or BHB as alternative energy sources saves compromised glycogen and helps restore its levels during chronic treatment. Thus, restored glycogen levels can be taken as an overall indicator of normalized energy homeostasis in TG mice.
At present, the detailed mechanism of OES effects on the Aβ–energy metabolism interaction is yet unclear. There is some experimental evidence that the Aβ interaction with glucose metabolism mainly takes place in astrocytes (Allaman et al. 2006). Aβ activates glial nicotinamide adenine dinucleotide phosphate oxidase (NOX) (Abramov et al. 2004) that leads to the overactivation of PARP, consequent depletion of NAD(P)H in astrocytes (Abeti et al. 2011), and finally to a reduction of astrocytic production of lactate, which is an important fuel for neurons during their activity (Pellerin and Magistretti 2011). Interestingly, this deleterious chain of effects is largely interrupted by the exogenous oxidative energy substrates such as pyruvate (Abeti et al. 2011). In addition, oxidative stress resulting from Aβ action may induce an increase in intracellular Cl− concentration in neurons (Sah and Schwartz-Bloom 1999), the effect similar to that we have observed in our experiments.
Without doubt, in this initial stage of investigation, we have only approached ‘the tip of an iceberg’ and a number of important issues remain to be addressed in future research. For instance, it is yet unclear why the normalization of brain energy supply prevents the toxic Aβ action. One explanation could be based on the hypothesis that energy metabolism is the primary Aβ target. It is yet unknown how OES would affect AD pathogenesis or whether OES could facilitate symptoms at later stages of the pathology. In this study, we mostly targeted neuronal excitability as a very early indicator of AD pathology (Bakker et al. Dickerson et al. 2005; Quiroz et al. 2012; Filippini et al. 2009). Indeed, behavioral tests would not be relevant for such context as the signs of cognitive impairments for this AD model have not been observed before 10 months of age (Minkeviciene et al. 2008). However, we have previously reported significant progressive impairment in spike timing-dependent plasticity in the very same TG mice that were several months younger (Shemer et al. 2006). In this study, we show that the acute Aβ1-42 application strongly reduces the tetanus stimulation-induced LTP, while this toxic effect is completely prevented by the presence of OES. Thereafter, the OES action should be verified on the TG animals at the age when cognitive impairment is already evident. Clarification of these critically important issues is a matter of future studies.
Considering OES administration to humans, limited studies suggest that it can improve cognitive function in mild to moderate AD (Henderson 2008; Henderson et al. 2009). Oral hydroxybutyrate is considered safe and well tolerated (Vernia et al. 2012). However, the high cost of pure BHB makes the use of the alternative energy substrate, pyruvate, more feasible for practical applications. Pyruvate is widely in use as an inexpensive dietary supplement because of its anabolic properties. It was shown to increase lean body mass while decreasing adiposity in experimental animals and humans [for review, (Egras et al. 2011)], thus demonstrating the ability to modify energy metabolism pathways. Neuroprotective effects of pyruvate have been reported in cases of brain ischemia, hypoglycemia, hemorrhagia, stroke, and kainate-induced epileptic brain damage [for review, (Zilberter et al. 2010)]. Recent studies have shown the therapeutic potential of pyruvate in the case of Leigh syndrome (Komaki et al. 2010; Koga et al. 2011).
Altogether, our results support the metabolic hypothesis of AD pathology (Blass 2007), postulating metabolic crisis as a major risk factor. We suggest that the vicious Aβ–energy metabolism cycle can be interrupted by the OES-induced correction of energy metabolism, potentially preventing/abating early AD pathogenesis. Our findings may open a new avenue for AD treatment, a strategy promising to be not only efficient but also free of side effects characteristic for the chronic administration of xenobiotic substances.
We thank Drs D. Borchelt (Univ. South Florida, FL, USA) and J. Jankowsky (Baylor College of Medicine, Houston, TX, USA) for providing the breeder APPswe/PS1dE9 mice. We thank Drs. C. Bernard and D. Turner for critically reviewing the manuscript. We thank Dr. T. Zilberter for valuable participation in the manuscript preparation. We thank Ms. Laila Kaskela for technical assistance in ELISA assays. This study was supported by the European Union Seventh Framework “MEMOLOAD” grant (HEALTH-F2-2007-201159; M.Z., H.T., T. H, Y. Z.), French National Research vAgency “METANEX” grant (ANR-2010-BLAN-1443-01; A. I. and Y. Z.), Alzheimer's Association International Research Grant (A.I. and Y.Z.), Academy of Finland (A.P., H.T.), The Sigrid Juselius Foundation (A.P.), Finnish Ministry of Education grant (S.Z), Epilepsy Foundation (Finland, S.Z.). The authors have no conflicts of interest to declare.