|Home | About | Journals | Submit | Contact Us | Français|
Sleep is one of the most pervasive biological phenomena, but one whose function remains elusive. Although many theories of function, indirect evidence, and even common sense suggest sleep is needed for an increase in brain energy, brain energy levels have not been directly measured with modern technology. We here report that ATP levels, the energy currency of brain cells, show a surge in the initial hours of spontaneous sleep in wake-active but not in sleep-active brain regions of rat. The surge is dependent on sleep but not time of day, since preventing sleep by gentle handling of rats for 3 h or 6 h also prevents the surge in ATP. A significant positive correlation was observed between the surge in ATP and EEG NREM delta activity (0.5–4.5 Hz) during spontaneous sleep. Inducing sleep and delta activity by adenosine infusion into basal forebrain during the normally active dark period also increases ATP. Taken together, these observations suggest that the surge in ATP occurs when the neuronal activity is reduced, as occurs during sleep. The levels of phosphorylated AMP-activated protein kinase (P-AMPK), well known for its role in cellular energy sensing and regulation, and ATP show reciprocal changes. P-AMPK levels are lower during the sleep-induced ATP surge than during wake or sleep deprivation. Taken together, these results suggest that sleep-induced surge in ATP and the decrease in P-AMPK levels set the stage for increased anabolic processes during sleep and provides insight into the molecular events leading to the restorative biosynthetic processes occurring during sleep.
The subjective experience of sleep as restorative of energy is a commonsense observation, but one not directly studied physiologically with modern technology in discrete brain regions. The importance of sleep and suggestions about its physiological role have been better documented as a negative, by what happens without sleep, since prolonged sleep deprivation (SD) or sleep restriction adversely influences metabolic processes (Knutson, 2007), general emotional and physical health (Haack and Mullington, 2005), and neurocognitive behavior (Lim and Dinges, 2008). An often postulated, although not directly measured, function of sleep is to restore brain energy expended during active waking (Benington and Heller, 1995). Although constituting only 2% of body mass, brain oxygen and glucose utilization account for approximately 20% of those of the whole organism (Magistretti, 1999). Compared with wakefulness, indirect evidence that sleep reduces brain energy demands is a 44% reduction in the cerebral metabolic rate (CMR) of glucose (Maquet, 1995) and a 25% reduction in the CMR of O2 (Madsen et al., 1991). Our previous reports indirectly support a link between wake-related neural activation and energy expenditure, since felines showed an increase in extracellular levels of a metabolic byproduct of energy, adenosine, in a wake-active brain region, the basal forebrain (Porkka-Heiskanen et al., 1997) and a decline during spontaneous sleep, a pattern also observed in rodent basal forebrain (McKenna et al., 2003). Moreover, adenosine levels increase markedly if sleep is prevented (SD) (Basheer et al., 1999; Basheer et al., 2004).
These adenosine studies prompted us to examine the actual “currency of brain cellular energy” adenosine-triphosphate (ATP), since adenosine may be an indicator of neuronal activity-dependent energy use, by reflecting ATP breakdown. Steady-state ATP levels were once considered to be stable. However, recent brain studies indicate that electrical stimulation, glucose deprivation or manipulations of Na+/K+ATPase activity induce detectable changes in ATP levels (MacLean and Luo, 2004; Bao et al., 2005; Christian et al., 2008).
Here we report that ATP levels are maintained at a steady state levels during spontaneous waking but the levels exhibit a surge in the initial hours of sleep in brain regions with predominantly wake-active neuronal activity, a surge abolished by preventing sleep, whereas in the “sleep-active” ventrolateral preoptic (VLPO) region preventing sleep does not change ATP levels. In wake-active brain regions, the spontaneous sleep ATP surge positively correlates with the intensity of NREM delta activity (slow wave delta range 0.5–4.5 Hz), a marker of homeostatic sleep pressure. This ATP-delta correlation is also confirmed by pharmacologically induced sleep using adenosine perfusion into the basal forebrain. Furthermore, the changes in ATP exhibit reciprocity with the phosphorylated state of the cellular energy sensor, phosporylated adenosine-monophosphate (AMP) activated protein kinase (P-AMPK), thus, supporting the induction of anabolic processes during sleep.
Male Sprague-Dawley rats (350–400g) were used for the present study. The rats were housed in cages with a 12-hr light-dark cycle (lights on 7:00AM to 7:00PM, a constant temperature of 23°C and food and water ad libitum. Animals were treated in accordance with the Association for Assessment and Accreditation of Laboratory Animal Care and Use Committee at Boston VA Healthcare system, Harvard University and U.S. National Institute of Health. Every effort was made to minimize animal suffering and to reduce the number of animals used. For electroencephalogram (EEG) recordings the rats were implanted with EEG and electromyogram (EMG) electrodes under general anesthesia (i.m. ketamine 7.5mg/100g body weight, xylazine 0.38mg/100g, acepromazine 0.075mg/100g). EEG electrodes (stainless steel screws) were implanted epidurally over the frontal (primary motor, AP=+2.0; ML=2.0) and parietal (retrosplenial, AP=−4.0; ML=1.0) cortices. EMG recording electrodes (silver wires covered with Teflon) were implanted into neck muscles (Basheer et al., 1999).
In AD and CSF treated animals, intracerebral guide cannulas (CMA11guide, CMA Microdialysis, Stockholm, Sweden) were implanted 2 mm above the target. The target coordinates for the probe tip in the basal forebrain were AP, −0.3; ML, 2.2; DV, 8.8. After the surgery, the rats were housed in individual cages.
Seven days after surgery, the animals were transferred to the recording cages in a sound-attenuated room for habituation with attached EEG cables. Baseline EEG was monitored for a 24h period starting at 7am. The EEG/EMG signals were amplified and sampled at 104Hz. EEG recordings (acquisition using Grass Gamma v4.3) were scored using the Rat Sleep Stager (version 3.2) in 10sec epochs manually for NREM sleep, REM sleep and wakefulness. Recordings were divided into 1 hour bins for the 12 h sleep period (light period 7AM to 7PM); the amounts of wake, NREM sleep, REM sleep were calculated for the 12h light period. To determine the changes in slow-wave EEG power in delta range (0.5 – 4.5Hz), each animal’s mean delta power during NREM episodes for 12h light period was first measured. Then, for each hour’s NREM delta power, the % change from the mean was determined. For spontaneous sleep, delta power furnishes a reasonable proxy for neuronal activity, increased delta power implying overall decreased neuronal activity in wake-active neuronal regions (Steriade and McCarley, 2005; Vyazovskiy et al., 2009). However, in cortex, the negative delta-neuronal activity correlation breaks down for post-deprivation sleep, when neuronal activity greatly increases during the “on” states of cortical delta activity (Vyazovskiy et al., 2009).
For the microdialysis experiment, 10 rats, implanted with EEG/EMG electrodes and microdialysis cannula, were used. After habituation in the experiment room, 24 h EEG recordings (7AM–7AM) were done on the first day (baseline day). On the second day microdialysis probes were inserted into the target area via the guide cannula approximately 20 h before the start of the experiment. On the third day the microdialysis lines were connected to the probe and pump at 4PM and the artificial cerebrospinal fluid (aCSF: 147 mM NaCl, 3 mM KCl, 1.2 mM CaCl, 1.0 mM MgCl ; pH 6.6) perfusion was begun at 5PM. One group of rats (n = 5), indicated as AD group, was unilaterally perfused with aCSF for 2 h (5PM–7PM), followed by 3 h (7PM–10PM) of perfusion with adenosine (AD, 300μM, Sigma-Aldrich), whereas the other group, aCSF controls (n = 5), continued to receive aCSF. Previous studies from our laboratory showed that unilateral perfusion with 300μM AD increased NREM sleep and delta power, when compared to rats perfused with aCSF (Basheer et al., 1999). EEG was recorded throughout the whole infusion period. At the end of the experiment the rats were decapitated, the microdialysis probe removed and Frontal Cortex (FC), Basal Forebrain (BF), Cingulate Cortex (CCX) and Hippocampus (HIPP) dissected as described below. We note that the ipsilateral side was used for confirming the position of the tip of the cannula and the contra-lateral side was used for ATP analysis.
SD was done by “gentle handling” which, according to standard protocols (Franken et al., 1991) involved presentation of new objects into the cage or gentle touching by a brush or hands when new objects did not keep awake. SD either began at 7AM (ending at 10AM for 3 h SD or 1 PM for 6 h SD) or began at 10AM, ending at 1PM. During SD the rats continued to have access to food and water ad libitum. During this period rats consumed twice the usual amount of food compared to sleeping controls.
The rats were killed by decapitation and brains removed. Coronal slices (2 mm thick) were carefully placed on a dry-ice (−78.5°C) containing covered Petri-dish for rapid freezing and subsequent dissection. Six brain regions were dissected: FC (tissue vol. ~2mm x 2mm x 1 mm, Bregma 4.2 to 2.2), BF (~2mm x 1mm x 1mm, Bregma −0.26 to −1.2), CCX (~2mm X 2mm x 1 mm, Bregma −0.26 to −1.2), lateral hypothalamus (LH) ( ~1mm x 1.0mm x 0.2mm, Bregma −1.3 to −1.5), VLPO (~0.2mm x 0.2mm x 0.2mm, Bregma −0.3 to −0.5) and the entire HIPP. Extreme care was exercised to complete this process rapidly, with an average time of 80 + 9sec for tissue collection, both during dark and light periods for all the animals; collection times did not differ between dark and light periods. The dissected regions were kept frozen on dry ice and stored at −80° C until used for biochemical measurements.
Determination of ATP was performed by a luciferin-luciferase based Assay (Lundin and Thore, 1975; McElroy and DeLuca, 1983) using a commercial ATP Assay system with bioluminescence detection kit (Enliten, Promega). The assay principle is that, in the presence of ATP and oxygen, luciferase from Photinus pyralis catalyses D-luciferin to oxyluciferin, Pi, AMP, carbon dioxide, and light. The light intensity is measured by luminometry. This technique has been widely used for ATP measurement in cell cultures, slices and also to measure in vivo changes using frozen dissected tissue including brain tissue (Bao et al., 2005; Park et al., 2006; Chang et al., 2008; Christian et al., 2008). ATP was measured according to the manufacturer’s protocol. Briefly, weighed tissue samples were homogenized in 5% trichloroacetic acid (TCA) and transferred to 1.5ml Eppendorf tubes. The samples were centrifuged at 5000 rpm in cold for 5 min and the supernatant were transferred to a fresh tube. Samples (10μl) were neutralized with Tris Acetate buffer (490μl) adjusted to a pH value of 7.75. The luciferase reagent was added immediately before measurement in the luminometer (Flexstation III, Molecular Devices, Sunnyvale, CA), as described by the supplier. A new standard curve was made daily before each measurement using known standards and ATP-free water. The absolute concentration was calculated per mg wet tissue weight using known ATP standards provided in the kit. ATP concentrations are expressed as 10−8 M/mg tissue. This method gave highly reproducible results. The mean Coefficient of Variation (standard deviation/mean) at the same time of day 10 PM in the brain regions was 0.13.
The concentrations of phosphocreatine (PCr), creatine (Cr) and AMP were measured using reversed phase HPLC and UV-detection as described earlier (Helzberg et al., 1987; Dworak et al., 2007). A C-18, 150 x 3.9 mm, 4μm particle size column (Waters), was equilibrated with a mobile phase (buffer A) containing 14mM H2KO4P and 3mM tetrabutylammonium bisulfate (Sigma) adjusted to pH 5.4 with 2M KOH. A step gradient was obtained with the second mobile phase (Buffer B) containing 70% methanol. The gradient was formed as follows: 0–4 min 100% Buffer A; 4–20 min up to 40% buffer B; 20–25 min 40% buffer B; 25–35 min up to 100% buffer A. The flow rate throughout chromatographic runs was 1.0 ml/min. A new standard curve was made daily. All reagents were of the highest purity available. PCr and Cr were purchased from Fluka. AMP was purchased from Sigma-Aldrich. TCA-supernatants were neutralized with 2.0 M K2HPO4. The compounds (PCr, Cr, AMP) were identified on the basis of the retention time. Quantitative analysis of both standards and samples was performed at 210 nm wavelengths for PCr and Cr and 254nm wavelength for AMP.
The changes in P-AMPK levels were examined in frontal cortex and basal forebrain only. These two regions exhibit the highest surge in ATP during sleep. In BF, tissue samples (n = 4 each group/region) were collected at 7AM (baseline ATP levels), 10AM (highest surge in ATP), 1PM and 7PM. Samples were also collected and after 3 h of SD (7AM–10AM, no surge in ATP, levels equal to baseline levels), and after 3 h SD followed by 3 h RS (ATP levels rise with recovery sleep). In FC the samples were collected at 7AM, 10 AM and following 3 h of SD. The samples were collected with extreme rapidity on dry ice as described above. Samples were homogenized on ice in a Lysis-Buffer containing a RIPA buffer, okadaic acid (25 μM), Staurosporine (25μM) and a protease inhibitor and transferred to an eppendorf tube. Samples were centrifuge at 5k in cold for 5 min and the supernatant transferred to a fresh eppendorf tube. Protein concentration was estimated using BCA methods (Pierce Tech.) and 20 μg protein samples were electrophoresed in 12% SDS-PAGE gels. Proteins were electroblotted onto nylon membranes for Western blotting. P-AMPKα1 rabbit polyclonal antibody (1:1000 dilution, sc-33524, Santa Cruz Biotechnology, Inc.) was used for the detection of Thr172 P-AMPKα1 by Western Blotting. HRP conjugated goat anti-rabbit (1:5000) was used to detect the western blot using Western Lightening chemiluminescence Reagent Plus (Perkin Elmer). The membranes were further probed with antibody against β-actin and the band densities for P-AMPK were normalized using β-actin. We did not repeatedly freeze/thaw samples, since previous work found that P-AMPK levels increase in homogenates because of freeze/thawing (Scharf et al., 2008b).
Male Sprague Dawley rats were maintained under 12 h light/12 h dark periods. Every three hours for 24 hours, starting from 7AM (lights on, onset of normal sleep period), ATP levels were determined using a validated luciferin-luciferase ATP detection assay in FC, BF, CCX, and HIPP. Within each of the four brain regions, the steady-state ATP level was stable during the wake period (7PM–7AM, lights off), but ATP levels dramatically altered values in all four regions during the sleep period (7AM–7PM, lights on). The average ATP molar values during the wake (dark) period calculated per mg tissue wet weight varied between regions (n = 6), but were relatively stable within a region. In contrast, during the light period, when rats were asleep most of the time (66.7 + 6.3%), ATP levels in each brain region generally were elevated and, most importantly, were not constant, but showed significant alterations in values within each brain region as compared with waking (Figure 1).
The increase in ATP during sleep followed a distinct pattern. The lowest levels were seen at 7AM and were used as a baseline for comparison of light (sleep) period values and were comparable to the average ATP levels during the wake-period (7PM–7AM). In the initial sleep period in all four brain regions, ATP levels surged significantly, showing highest values at 10AM (n = 6; P < 0.01) that declined slightly by 1PM, although they remained significantly higher than the baseline values (n = 6; P < 0.01) (Figure 1). Using the 7AM values as baseline for diurnal comparisons, the percentage increases at 10AM and 1PM were comparable in FC (265 + 65% and 214 + 67%) and BF (235 + 49% and 174 + 28%), but greater than the increases noted in CCX (114% + 35 and104 + 11 %) and HIPP (90 + 20% and 29 + 10%) (Figure 1). The levels declined to baseline values by 4PM. This initial ATP surge follows the initial sleep period surge of EEG NREM delta activity (delta activity, 0.5–4.5 Hz). Delta activity, like ATP, showed significant variability (n = 5, P < 0.013) during sleep, declining toward the end of the sleep period (Figure 2A and 2B).
To furnish an experimental manipulation of EEG NREM delta activity that would further test our hypothesis of NREM delta-ATP association and to rule out potentially confounding diurnal effects, such as light, we used microdialysis to perfuse adenosine (300μM) unilaterally into the BF at the onset of the active (dark) period, from 7PM to 10PM, a procedure previously shown by us to increase NREM delta (Basheer et al., 1999). Indeed, during microdialysis perfusion, compared to the rats that were perfused with aCSF, adenosine perfused animals showed higher increases in % time in NREM sleep and NREM delta activity when compared to the same time period on a baseline day with no perfusion (Figure 3A). There was a significant increase in ATP at the end of adenosine perfusion in all 4 brain regions compared with aCSF controls (Figure 3B), whereas AMP concentrations showed a tendency towards decrease with statistically significant decrease in BF only (FC: −8.91±3.99% (p=0.137); BF: −18.19±6.01% (p=0.032); CCX: −13.18±8.54% (p=0.343); HIPP: −18.34±7.90% (p=0.072); t-test). An association between NREM delta and ATP over the 3 h of perfusion was observed, as shown by their strong correlation in FC (rho = 0.830, P = 0.0006), the site of EEG recording, and BF (rho = 0.636, P = 0.04) a site known to be related to delta activity (Basheer et al., 1999) (Figure 3C and 3D).
To further distinguish whether the surge in ATP was associated with the time of day (diurnal) variation or with sleep behavior, we subjected rats to 3 h SD by gentle handling starting at 7AM, and examined the ATP levels at 10AM, the time when ATP levels were highest in diurnal control animals. This SD blocked the ATP surge seen in controls in each of the four brain regions in the SD animals (Figure 4A, B, C, and D). When the rats were allowed 3 h recovery sleep (RS) after the 3 h of SD, only in BF did the ATP levels surge to match the levels of the diurnal controls (1PM). In the other brain regions, 6 h of RS was needed to induce the surge in ATP levels. Thus, SD for 3 h postponed the ATP surge in all four regions, although with variable time lags. Interestingly, ATP levels did not decline below the baseline (7AM values), suggesting a balance between the ATP synthesis and usage during short-duration SD.
Next we addressed whether interrupting the sleep period would also influence the pattern of ATP change observed in normally sleeping animals. Rats were allowed to sleep during the first 3 h of the light period (7–10AM), allowing the initial ATP surge to occur, and then received 3 h SD (10AM–1 PM). In normally sleeping control animals, the ATP levels during this period begin to exhibit a slow decline, although still considerably higher than the baseline (Figure 1). SD at this point in the sleep period increased the rate of decline in ATP, so that ATP levels reached baseline by the end of the 3 h SD, while this transition took 6 h in the normally sleeping animals. ATP levels after 3 h SD were significantly lower in all four regions (BF, P = 0.022; FC, P = 0.006; CCX, P = 0.004; HIPP, P = 0.004, n = 6/region) compared to their respective diurnal controls (Figure 4E, F, G, H). When 3 h SD was followed by 3 h of RS (1PM–4PM), ATP levels in all brain regions increased during the RS to levels higher than those in their diurnal controls (4PM), again demonstrating the sleep dependence of the ATP increase (Figure 4E, F, G, H).
We then examined the effect of a longer duration of SD (6 h, 7AM–1PM) on ATP. In contrast to 3 h SD, 6 h SD produced a small but statistically significant decrease below the baseline (7AM) level in all brain regions (each P < 0.05) except for FC, where the levels matched the baseline (Figure 4I, J, K, L). At 4PM, following 6 h SD and 3 h RS, ATP levels in all four brain regions were comparable to the diurnal control values, which were near the baseline levels. Taken together, these results indicate that during waking (either spontaneous or due to SD) ATP levels remain close to the 7AM baseline values (3 h SD) or even slightly lower (6 h SD), while sleep onset results in a rapid and significant surge in ATP levels. These effects of sleep and wake behavior on brain ATP levels are thus independent of the time of day.
To determine if the increase in ATP is related to local neuronal activity patterns associated with sleep and wake, in a separate experiment (n = 5/group) we compared the 3h SD-induced changes in the levels of ATP in two functionally diverse regions of hypothalamus, namely the LH known to predominantly contain wake- and REM- active neurons (Szymusiak and McGinty, 2008; Hassani et al., 2009), and the VLPO that predominantly contains sleep-active neurons (Sherin et al., 1996). In the same rats we also re-examined FC. After 3h SD, ATP concentrations were significantly reduced in FC (−53.36±8.01% (p=0.007)) when compared to sleeping controls as was observed in the previous group of rats. ATP also showed significant decrease in LH (−40.17±19.8% (p=0.048)), whereas no significant change was observed in the sleep-active VLPO (+10.69±20.36% (p = 0.719)), (Figure 5).
To further examine the SD-induced changes in the energy status we measured the levels of PCr and Cr in FC, BF, CCX and HIPP. PCr serves as a rapidly mobilizable reserve of high-energy phosphates in brain. In the event of increased ATP use, PCr can anaerobically donate a phosphate group to ADP to form ATP. The levels of PCr showed a trend increase at 10AM when ATP levels surge. The levels of PCr decreased significantly in FC (−57+17%, P<0.001), BF (−58+23%, P<0.01), and CCX (−45+20%, P = 0.044) following 3 h SD (7AM–10AM) when compared to undisturbed time matched (10AM) controls. HIPP did not show significant change (−24+27, P =0.228). The Cr levels showed a significant increase in HIPP (+42+5%, P=0.002) whereas other three regions showed trend increases (Figure 6).
We next addressed the mechanism causing the sleep-associated initial surge in ATP to return to baseline levels during the later hours in the light (sleep) period. We investigated whether the AMPK has a role in detecting and responding to ATP changes during sleep and SD. This phylogenetically-conserved kinase monitors changes in cellular concentrations of ATP and AMP. Increased ATP usage (higher AMP/ATP ratio) activates AMPK by preventing the dephosphorylation of this kinase (Hardie, 2007). P-AMPK, in turn, phosphorylates multiple downstream target proteins, regulating cellular energy metabolism by inhibition of ATP-consuming anabolic pathways and activation of ATP-generating catabolic pathways (Hardie, 2007). We hypothesized that at 7AM, when the AMP/ATP ratio is presumably high after the 12 h of predominantly active period, a more elevated level of P-AMPK would be detected than at 10AM, at the peak of the sleep-induced ATP surge, accompanied by a decreased AMP/ATP ratio. In contrast, if sleep were to be prevented by 3 h SD between 7AM-10AM, P-AMPK levels would be expected to remain high until recovery sleep is allowed.
To test this hypothesis, first we examined the changes in P-AMPK levels in BF at 7AM, 10AM, 1PM and 7PM during lights on period and compared them with 3h of SD (10AM) followed by 3h of RS (1PM). As shown in Figure 7A and B, the levels of P-AMPK decrease at 10AM (n = 4, P = 0.02) and 1PM and show a tend increase at 7PM. If sleep is prevented for 3h, the level of P-AMPK increases significantly (n = 4, P = 0.01) at 10AM when compared to the undisturbed sleeping time matched controls and the rats that were allowed 3 h recovery sleep following SD. To further confirm the inverse changes in P-AMPK and ATP, we compared the levels of P-AMPK in BF and FC at two diurnal time points, 7AM and 10AM with that of the changes in ATP levels at the same time point (ATP levels are highest at 10AM when compared to 7AM baseline). As was the case with BF, a reciprocal relationship between P-AMPK and ATP was observed in FC (trend-level, 7AM vs 10AM, P = 0.13). Also in accord with this hypothesis, similar reciprocal relationships between P-AMPK and ATP levels were also observed when 3 h SD rats were compared with 10AM diurnal controls (n = 4, BF, P < 0.02; FC showed a similar trend, P < 0.08) (Figure 7C, D, E, and F).
Our results provide molecular evidence that ATP, the energy currency of brain cells, shows a surge in the initial hours of sleep which is tightly correlated with EEG NREM delta activity during spontaneous sleep, and is accompanied by reciprocal changes in phosphorylated AMPK, a main regulator of anabolic and catabolic pathways. Our results are consistent with earlier work on whole brain energy metabolism during sleep-wake, showing an overall increase of high-energy phosphates during sleep (Van den Noort and Brine, 1970; Durie, 1978; Dworak et al., 2007). However, early work on brain energetics presented contradictory views on energy changes during sleep (Reich et al., 1972; Durie, 1978). Moreover, these studies did not parcellate brain into individual regions, did not measure ATP during a 24h light-dark cycle, and did not record sufficient electroencephalographic activity to distinguish wake, NREM and REM sleep. Thus, to our knowledge, this is the first report of ATP changes during sleep in discrete brain regions and its direct relationship to NREM delta activity and P-AMPK.
The diurnal pattern of ATP rise and fall is strikingly similar to the rise and fall of slow wave delta activity, a marker of sleep homeostasis (Borbely, 2001) during the sleep period (Figures 1, 2A and 2B). A similar profile of slow wave delta activity during the sleep period was also described by Dash et al., (Dash et al., 2009). Preventing sleep by keeping the rats awake by gentle handling (SD) prevents the surge in ATP. During recovery sleep that followed SD, the apparent association between the ATP surge and slow wave delta activity observed during spontaneous sleep, was limited to BF only and not seen for cortical ATP which increased only after 6 h of RS (Figure 4A,B, C, and D). BF has been implicated in RS delta activity (Kalinchuk et al., 2008; Kaur et al., 2008). Our previous work has implicated the adenosine A1 receptor in post-synaptic inhibition of wake-active neurons and the promotion of delta activity (Rainnie et al., 1994; Arrigoni et al., 2006). Importantly, the inhibitory effects of adenosine, generated from ATP released by gliotransmisson, acting on the A1 receptor have also been shown to play a role in both the delta activity and the slow oscillation (<1Hz) of NREM sleep, and thus gliotransmission may contribute to our ATP findings (Fellin et al., 2009; Halassa et al., 2009). Thus, the net surge in ATP levels during the initial period of sleep is likely attributable to a decreased brain energy use during NREM, when neuronal activity and accompanied energy consumption are low.
While it is known that SD evokes an increase in delta activity during initial hours of subsequent recovery slow wave sleep that is proportional to the sleep loss (Tobler and Borbely, 1986; Steriade and McCarley, 2005), the uncoupling between cortical ATP surge and recovery sleep delta (data not shown) that occurs in the initial hours of RS is intriguing. The absence of ATP surge in cortex suggests continued ATP utilization during initial periods of RS. Indeed the findings of Vyazovskiy et al., (2009) indicate that, although delta is increased in RS following SD, there is a simultaneous increase in neuronal activity in cortex, due the increased firing during the “on” states of delta activity compared with non-deprived control animals. Thus, while post-deprivation delta is an imperfect indicator of neuronal activity, this direct measurement of neuronal activity supports our main contention that decreased neuronal activity promotes ATP increases and increased activity leads to ATP decreases.
A direct relationship between neuronal activity and energy consumption was suggested as early as 1890 by Roy and Sherrington and confirmed by numerous studies (Attwell and Laughlin, 2001; Dhar and Wong-Riley, 2009; Magistretti, 2009). More recent work indicates that some 87% of brain energy consumption is proportional to neuronal firing rates, predominantly consumed in the restoration of membrane potential after postsynaptic potentials, while 13% is used to maintain resting potentials, indicating that excitatory neurotransmission and Na+-K+-ATPase activity consumes most of the energy expended by the brain ( Alle et al., 2009; Magistretti, 2009). In accordance with this is our observation on VLPO, an area known to predominantly contain sleep-active neurons (Sherin et al., 1996), where SD did not have any effect on its ATP levels (Figure 5). On the contrary, in other wake active brain areas SD-induced continued neuronal activation resulted in significant decrease in phosphocreatine levels (Figure 6) suggesting rapid mobilization of high-energy phosphates from phosphocreatine in order to prevent ATP depletion.
The observed high ATP levels during sleep might have multiple functional consequences. ATP acts as an activity-dependent metabolite and signaling molecule between neurons and glia and modifies membrane potentials through ATP-modulated potassium (KATP) channels (Fields and Stevens, 2000; Peters et al., 2004; Pascual et al., 2005) thus strongly influencing neuronal information processing, synaptic strength, gene expression, protein, fatty acid, and glycogen synthesis (Fields and Stevens, 2000; Peters et al., 2004; Pascual et al., 2005). One of the supporting evidence is from the data showing increased glycogen biosynthesis during sleep (Karnovsky et al., 1983; Kong et al., 2002; Petit et al., 2002). However, other studies showed that SD did not affect brain glycogen (Franken et al., 2006) and that the changes are dependent on the age and strain of the animals (Gip et al., 2002). Due to these variable results, and its limited role in total brain energy contribution (Gruetter, 2003) the significance of glycogen metabolism during sleep is not clear.
High ATP levels would also provide the necessary energy for subsequent plasticity changes, (Hoeffer and Klann, 2010; Dennis et al., 2001). Higher energy utilization during wake and its reversal during sleep is also congruent with a current hypothesis of synaptic homeostasis (Tononi and Cirelli, 2006), since increased synaptic potentiation (synaptic weight) during wake would contribute to high energy consumption and the surge in ATP during sleep would reflect the energy conserving processes such as synaptic downscaling during sleep. Also of note, the availability of ATP in sleep could promote the changes related to synaptic plasticity and memory consolidation (Huber et al., 2004; Stickgold, 2005; Huber et al., 2006; Landsness et al., 2009).
Our results provide direct evidence for an energy-dependent regulation of sleep-related anabolic processes by the reciprocal relationship of P-AMPK and ATP. We observed that P-AMPK levels are high during wake and decrease during sleep whereas SD induce a significant increase in P-AMPK (Figure 7). AMPK is known as a major sensor and regulator of cellular energy status through the regulation of multiple biochemical pathways (Hardie, 2007). The phosphoryaltion status of this protein is changes by the in the intracellular AMP/ATP ratios. During conditions of high ATP use (high AMP/ATP ratio) as observed during spontaneous waking and SD, AMPK is phosphorylated. Higher P-AMPK switch on catabolic processes that provide alternative routes to generate ATP, such as glycolysis, fatty acid oxidation and mitochondrial biogenesis, while switching off ATP-consuming anabolic processes (Hardie, 2007). In contrast, when ATP utilization is at its minimum (low AMP/ATP ratio) as observed during sleep, decreased levels of P-AMPK promote anabolic processes via multiple downstream pathways, including the synthesis of fatty acids, glycogen and protein (Hardie, 2007). P-AMPK regulates protein synthesis at multiple points, including the mTOR pathway or by activating eukaryote elongation factor 2 (eEF2) (Hardie, 2007). Increased P-AMPK represses the mTOR signal cascades and prevents the maintenance of Late-Phase Long-Term-Potentiation (LP-LTP) (Potter et al., 2010), thus directly linking energy metabolism and synaptic plasticity in the mammalian brain (Potter et al., 2010). A decrease in P-AMPK observed at 10AM when the ATP levels are high is suggestive of potentiation of anabolic process during sleep. This is also consistent with the recent findings on the role of sleep in memory consolidation, synaptic plasticity and remodeling. Our findings of an increased P-AMPK during waking and SD and a decrease in P-AMPK during sleep are in agreement with previous reports demonstrating an increase in the transcription of genes involved in protein synthesis and synaptic plasticity (Cirelli et al., 2004; Mackiewicz et al., 2007), increased translation of proteins during sleep (Ramm and Smith, 1990; Nakanishi et al., 1997), and, following SD, an overall decrease in protein synthesis (O’Hara et al., 2007).
In summary, our data showing an increase in ATP levels during sleep provide molecular evidence in support of the long-standing view that an important function of sleep is related to providing the brain with increased energy stores (Benington and Heller, 1995; Scharf et al., 2008a). Our data, however, significantly recast the sleep and energy restoration hypothesis. Instead of speaking of energy “restoration”, since ATP levels at the end of the wake period are not strikingly lower than at wake period onset, we restate the hypothesis as “sleep is for an energy surge”, a surge that permits energy-consuming anabolic processes, such as protein and fatty acid synthesis, to occur. Short-term SD delays and longer-term SD both delays and limits the extent of this ATP surge, and this decrease in the surge may impair energy-requiring biosynthetic processes. We summarize our hypotheses and data about the association between sleep-wake dependent changes in ATP, AMPK and accompanied anabolic and catabolic pathways in a figure (Figure 8).
In conclusion, our data suggest that an initial ATP surge nourishes the anabolic, restorative biosynthetic processes occurring during sleep, in accord with Shakespeare’s intuitive phrasing, “Sleep… great nature’s second course, Chief nourisher in life’s feast” (Macbeth, Act II, Scene II).
We gratefully acknowledge Farzana Pervin Nipa for technical assistance, and Diane Ghera and Dewayne Williams for help with animal care. This work was supported by awards from the Department of Veterans Affairs Medical Research Service Award to RB, a Deutsche Forschungsgemeinschaft fellowship (DW66/1-1) to M.D., and the National Institute of Mental Health (NIMH39683) to RWM.