Rapid changes in glutamate levels in the posterior hypothalamus across sleep-
wake states in freely behaving rats
Joshi John, Lalini Ramanathan and Jerome M Siegel*
Neurobiology Research (151A3), Veterans Affairs Greater Los Angeles Health Care System, North
Hills, CA 91343, U.S.A. and Department of Psychiatry and Brain Research Institute, University of
California at Los Angeles, Los Angeles, CA 91020
Running title: Real-time glutamate release in waking & REM sleep
Number of pages: 29
No. of Figures: 9; No. of Tables: 1
*To whom correspondence should be addressed
Jerome M. Siegel, Ph.D.
UCLA/Neurobiology Research (151A3)
16111 Plummer St.
16111 Plummer Street
North Hills, CA 91343
Articles in PresS. Am J Physiol Regul Integr Comp Physiol (September 24, 2008). doi:10.1152/ajpregu.90541.2008
Copyright © 2008 by the American Physiological Society.
The histamine containing posterior hypothalamic region (PH-TMN) plays a key role in sleep-wake
regulation. We investigated rapid changes in glutamate release in the PH-TMN across the sleep-wake
cycle with a glutamate biosensor that allows the measurement of glutamate levels at 1-4 sec resolution.
In the PH-TMN, glutamate levels increased in active waking (AW) and REM sleep compared to quiet
waking and NREM sleep. There was a rapid (0.6 + 1.8 sec) and progressive increase in glutamate
levels at REM sleep onset. A reduction in glutamate levels consistently preceded the offset of REM
sleep by 8 ± 3 sec. Short duration sleep deprivation resulted in a progressive increase in glutamate
levels in the PH-TMN, perifornical-lateral hypothalamus (PF-LH) and cortex. We found that in the
PF-LH, glutamate levels took a longer time to return to basal values compared to the time it took for
glutamate levels to increase to peak values during AW onset. This is in contrast to other regions we
studied in which the return to baseline values after AW was quicker than their rise with waking onset.
In summary, we demonstrated an increase in glutamate levels in the PH-TMN with REM/AW onset
and a drop in glutamate levels prior to the offset of REM. High temporal resolution measurement of
glutamate levels reveals dynamic changes in release linked to the initiation and termination of REM
Key words: REM sleep, histamine, microdialysis, biosensor, cortex
Various studies have implicated glutamatergic mechanisms in the control of sleep-wake states. In vitro
experiments reported a differential glutamate release in brain regions of long sleep and short sleep
mice (11). Glutamate release was higher during wakefulness than during sleep in the
pedunculopontine tegmental nucleus (22), whereas enhanced glutamate release during REM sleep was
observed in the rostromedial medulla (24). Paradoxical sleep deprivation increased the content of
glutamate and glutamine in the rat cerebral cortex (3).
Glutamate excites both cortical and subcortical neurons and enhances the in vivo release of histamine
(HA) through presynaptic NMDA receptors in the anterior hypothalamus of the rat (32). Single cell
recording and microdialysis experiments demonstrate that HA cell activity is reduced in NREM sleep
and REM sleep compared to waking (7, 20). We hypothesized that changes in glutamate release in the
PH-TMN and in the terminal regions of HA neurons could affect HA release and sleep-wake states.
Dysfunction of the Hcrt system produces the sleep disorder, narcolepsy (35, 38, 44). Hcrt neurons
contain glutamate as a co-transmitter (1). Hcrt effects on sleep-wake state distribution are to some
extent mediated by the histamine system (17).
The short duration of sleep-wake transition and sleep-wake states make it difficult to accurately
measure the time course of transmitter release across each state using the standard microdialysis-
HPLC technique, where samples are collected over a 3-5 min interval. In the present study we used a
newly developed glutamate sensor to measure changes in glutamate levels across sleep-wake states
with a temporal resolution of 3-5 seconds (16). We monitored glutamate levels across the sleep–wake
cycle in the PH-TMN and compared it with changes in the PF-LH and cortical areas in freely moving
rats. A preliminary presentation of the data has been made (21)
Materials and Methods
Male Sprague-Dawley rats weighing between 300 to 400g were used for this study. All the procedures
were done in accordance with the National Research Council Guide for the Care and Use of
Laboratory Animals. Animal protocols were approved by the Institutional Animal Research
Committee (IACUC) of the University of California at Los Angeles and by the IACUC of the
Veterans Administration Greater Los Angeles Health Care System. The rats were kept on 12h:12h
light dark cycle (6AM lights-on: 6 PM lights-off), ambient temperature 24 ± 1°C, with food and water
available ad libitum. Rats were anesthetized with ketamine/xylazine (80:10 mg/kg, i.p). Electrodes for
the assessment of sleep-wake parameters were chronically implanted under aseptic conditions, as
described previously (18). In brief, bilateral stainless steel screw electrodes were threaded through the
skull over the frontal and parietal cortex to record electroencephalogram (EEG). One screw electrode
threaded through the midline of the frontal bone was used as ground. Teflon coated multistranded
stainless steel wires with 2 mm exposed at the tips (Cooner Wire, Chatsworth, CA) were placed in the
dorsal neck muscles to record electromyogram (EMG). Guide cannulae with stylets (MD-2250,
Bioanalytical System Inc., Indiana) were implanted 2 mm dorsal to the targets in the posterior
hypothalamus histaminergic tuberomammillary nucleus (PH-TMN; A -4.2, L 1.7, H 8.9 mm ventral to
dura); perifornical lateral hypothalamus (PF-LH; A -3.2, L 1.3, H 7.5 mm ventral to dura) and frontal
cortex (A 1.2, L 2.5, H 2 mm ventral to dura) for subsequent placement of glutamate sensors (33).
Glutamate sensor data collection: One week after surgery, the rats were habituated to the recording
chamber and sleep-wake states were monitored. A pre-calibrated glutamate sensor with 1-mm long
active area was inserted through the guide cannula into the PH-TMN, PF-LH or frontal cortex and
fixed firmly to the guide cannula. The sensor was connected to a 4-channel potentiostat (model 3104,
Pinnacle Technology Inc., Kansas) using a flexible cable. The potentiostat was then connected to a PC
using a USB cable. PAL V1.2.6 software was used (Pinnacle Technology Inc., Kansas) for acquiring
sensor data. Data collection started 2 h after insertion of the glutamate sensor into the brain, at which
point a stable baseline was observed. Sleep-wake parameters and glutamate levels were recorded
simultaneously (from 11 AM- 6 PM, during the lights on period) in freely moving rats. Data was
collected during spontaneous active waking (AW), quiet wake (QW), non-rapid eye movement
(NREM) sleep, rapid eye movement (REM) sleep and sleep deprivation. Sleep deprivation studies
were performed on the second day. We used only sensor readings taken from the first 2 days after
implanting the sensor. We found that beyond the second day, the sensitivity of the sensor dropped
substantially in many cases.
An experiment was conducted to compare glutamate changes recorded using the glutamate sensor with
glutamate release using microdialysis-HPLC, across sleep-wake states. In three rats, along with the
glutamate sensor cannulae (21-G guide, non-steel) cannulae for microdialysis probes (20-G guide,
stainless steel) were also implanted into the PH-TMN. The microdialysis guide cannula (blocked with
a stylet) was implanted 0.4-0.6 mm lateral (center to center) to the glutamate sensor cannula.
Microdialysis protocol: After seven days of post-operative recovery, the rats with microdialysis
cannulae were habituated to the rodent microdialysis bowl (Bioanalytical Systems Inc. West Lafayette,
IN), and to the flexible EEG and EMG recording cables connected to commutators (Plastics One Inc.,
Roanoke, VA) which were connected to a Grass polygraph. The day before the recording session (16 h
prior to the experiment) the stylets were removed and a microdialysis probe was placed through the
guide cannula. Microdialysis probes (AI-01 Eicom, Japan; molecular cut-off size of 50 kDa) with a
semi-permeable membrane tip length of 1 mm and an outer diameter of 0.22 mm were used. The
length of the probe was set so that the glutamate sensor was 0.4-0.6 mm lateral to the semi-permeable
membrane and the estimated dialysis field of the probe included the extracellular environment of the
glutamate sensor. The microdialysis probe and glutamate sensor were fixed firmly (dental cement) to
minimize tissue trauma and to ensure maximum stability of recording. On the day of the experiment, a
glutamate sensor was inserted through the guide cannula into the PH-TMN and fixed to the cannula.
Then the adjoining dialysis probe was continuously perfused with artificial cerebrospinal fluid (aCSF;
145mM NaCl, 2.7mM KCl, 1.0mM MgCl2, 1.2mM CaCl2, and 2.0mM Na2HPO4, pH 7.4) at a flow
rate of 2 µl/min by a microinjection pump (Bioanalytical systems, West Lafayette, IN). The time taken
for the aCSF solution to travel from the tube to the tip of the microdialysis probe was previously
Microdialysis and glutamate sensor recordings: Data collection started 2h after insertion of the
glutamate sensor into the brain, when a stable baseline was observed. Microdialysis samples were
collected simultaneously at five-minute intervals (10 µl) using a refrigerated fraction collector (Eicom
Corp., Kyoto, Japan). Earlier published studies have reported that two hours of aCSF perfusion is
optimal for achieving a stable measurement of neurotransmitter release (19, 20). The polyethylene
collection vials were maintained at 5 °C during collection and then stored at -80°C until analyzed. The
recovery rate of the microdialysis probes for glutamate was 10-14%. A minimum of three
microdialysis samples for each sleep-wake episode were collected.
HPLC assay of glutamate: The concentration of glutamate in the dialysate was detected by HPLC (EP-
300, Eicom Corp., Kyoto, Japan) with fluorescent detection (Soma S-3350:
excitation/emission=340/440 nm), as described in our prior studies (19, 20, 28). The sample was
injected using an autoinjector (ESA 540, Eicom, Japan). Pre-column derivatization was performed
with o-phthaldialdehyde/2-mercaptoethanol at 5°C for 3 min. The derivatives were then separated in a
liquid chromatography column (MA-5ODS, 2.1x 150mm, EICOM, Japan) at 30°C with 30% methanol
in 0.1 M phosphate buffer (pH 6.0), degassed by an online degasser (DG-100, Eicom Corp., Kyoto,
Japan). Quantification was achieved with a PowerChrom analysis system (AD Instruments, Australia)
using external amino acid standards (Sigma, USA). Neurotransmitter concentrations were calculated
by comparing the HPLC peak of glutamate in the microdialysis samples with peak areas of known
concentrations of glutamate analyzed on the same day. The detection limit for glutamate is 20 fmol.
Recording of sleep-wake stages: Signals for the assessment of different sleep-wake stages (EEG and
EMG) were amplified with a Grass model 78E Polygraph (Grass Instruments, Quincy, MA). All
recordings were carried out between 11 AM- 6 PM, during the lights on period. EEG and EMG signals
were digitized using a CED 1401 Plus interface and analyzed manually using Spike2 software
(Cambridge Electronic Design Ltd., Cambridge, and UK). Sleep-wake stages were classified based on
electrophysiological parameters as previously described (18). In AW, the animal is alert with
movements, low voltage (desynchronized) EEG, increased levels of muscle tone and frequent eye
movements. In quiet wake (QW), the animal shows limited movement and often rests on the ground
with low voltage EEG and low EMG. NREM sleep is characterized by slow wave activity and low
EMG (further reduced from QW). REM sleep is characterized by low voltage EEG, muscle atonia and
Glutamate changes with sleep deprivation: Changes in glutamate levels before, during and after sleep
deprivation were recorded by glutamate sensors placed in the PH-TMN, PF-LH or cortex. EEG and
EMG recordings were also performed in this group of animals. Eleven rats (5 for PH-TMN, 3 for PF-
LH and 3 for cortex) were subjected to sleep deprivation starting at 11AM for periods ranging from
20-120 min. Sleep deprivation was performed by gentle handling which includes cage tapping, and
brushing the fur with a cotton tip applicator each time the animal showed signs of sleepiness. Sleep-
wake states were monitored by electrophysiological parameters and visual observation.
Histology: Rats were deeply anesthetized with pentobarbital sodium (60 mg/kg) and transcardial
perfusion was performed with normal saline followed by 4% buffered paraformaldehyde. After
treatment with 30% sucrose in 0.1 M phosphate buffered saline, the brains were cut at 40 μm intervals
using a microtome. Cresyl violet staining was done to verify placement of glutamate sensors and
We used a glutamate sensor (enzyme-based biosensor; Pinnacle Technology Inc., Kansas) for the
measurement of rapid changes in extracellular concentrations of brain glutamate in freely behaving
rats. This enzymatic glutamate biosensor (Platinum-Iridium electrode) with an integrated Ag/AgCl
reference electrode relies on the glutamate oxidase catalyzed conversion of glutamate and O2 to alpha-
ketoglutaric acid and hydrogen peroxide (H2O2; 16). The enzymatically produced H2O2 is detected by
its oxidation with an applied potential of 600 mV at Pt-Ir electrodes. Electroactive interferants present
in the brain which are oxidized at the same working potential are eliminated by means of a passive
selective membrane composed of nafion and cellulose acetate by co-immobilization of ascorbic acid
oxidase. This sensor allows the measurement of physiological extracellular concentrations of
glutamate with high sensitivity at low concentrations (micromolar). This electrode has sensing cavity
dimensions of 180-micron diameter and 1.0 mm length, with sensitivity of more than 3nA/10µM
glutamate. It has a fast response time (1-4 sec) allowing the monitoring of the physiological time
course of glutamate release.
Prior to implanting the glutamate sensor, an in-vitro pre-calibration was performed using glutamate
and ascorbic acid. Changes in glutamate levels were recorded as changes in electrical current. A linear
response to glutamate over the concentration range of 0-50µM was observed (Fig.1). Changes in
glutamate levels in the regions we studied were well within the detection limit of the glutamate
biosensor. We and other investigators, using various techniques, showed that glutamate levels across
different brain regions never exceeded 50 μM in rats (2, 15, 19, 28, 40).
Only those sensors that responded by at least 3nA/10µM glutamate and did not respond to ascorbic
acid (250 μM) were used for in-vivo experiments. A rapid (1-4 sec) response of the sensors varying
from 3.2 to 5.5 nA per 10µM glutamate was obtained. A post-experiment calibration was performed
immediately following removal of the sensor, after the in- vivo experiments. Four out of five AW and
REM sleep states, across all the brain regions studied showed consistent state related changes in
glutamate release using the glutamate biosensor.
Recordings of sleep-wake stages with Spike 2 software were started simultaneously with PAL
software for recording glutamate changes. Based on the calibration of each sensor (Fig. 1), we
converted PAL data (pA) to micromolar (μM) concentrations of glutamate during the corresponding
sleep-wake states. Six rats were used for PH-TMN and three rats each for cortex and PF-LH. The
changes in glutamate levels reported are the mean of pooled data from all AW, QW, NREM and REM
states for all rats.
After the completion of the experiment, histological examination of the brain sections verified the
location of the glutamate sensors and microdialysis probes in the PH-TMN and glutamate sensors in
the PF-LH and frontal cortex (Fig.2).
The time course of glutamate release was determined as the time taken from the ‘start’ to the ‘peak’
glutamate levels (as shown in Fig.3). The ‘start’ is defined as the first time point of glutamate increase
(more than 20 pA) from a stable baseline (less than 10 pA variation over 20 sec). The ‘peak’ is the
point of inflection at the beginning of the plateau (more than 20 pA variation at consecutive data
points over 20 sec). The time course of glutamate clearance was determined as the time taken from the
‘drop’ to the ‘post-basal’ glutamate levels. The ‘drop’ is the first time point of glutamate decrease
(more than 20 pA) at the end of the plateau. The ‘post-basal’ period is the first time point of glutamate
return to a stable post-basal reading lasting at least 20 sec. The duration of increased glutamate levels
was defined as the time taken from the ‘start’ to the ‘drop’ in glutamate levels.
Baseline glutamate concentration was taken as the average of the 20-sec prior to the ‘start’ of each
state dependent glutamate change. This reduces the effect of any circadian influence as well as any
gradual decline in the sensitivity of the sensor. The highest glutamate concentration was determined as
the average values from the ‘peak’ to the ‘drop’ in glutamate levels. The increase in glutamate
concentration was then calculated as the difference between the highest and the baseline readings.
The latency of glutamate changes relative to the onset of each sleep-wake state was calculated as the
time difference between the onset of the state, and the ‘start’ of increased glutamate levels. The
latency of glutamate changes relative to the offset of each sleep-wake state was calculated as the time
difference between the offset of the state, and the ‘drop’ in glutamate levels.
Posterior hypothalamus tuberomammillary nucleus (PH-TMN)
We observed a marked increase in glutamate levels in the PH-TMN coinciding with the onset of REM
sleep compared to NREM sleep (Fig.3A). The average increase in glutamate concentration in REM
sleep was 0.36±0.04 μM above that of the prior NREM sleep period (46 episodes, 6 rats). The start of
the increase in glutamate levels was tightly linked (0.6±1.8 sec) to the onset of REM episodes (i.e.
onset of EEG theta activity and muscle atonia, Figs.3, 4). The start of the drop in glutamate levels
preceded REM sleep offset by 8.0 ± 3.3sec (Figs.3, 4). The magnitude of the increase in glutamate
levels was not correlated with the duration of REM sleep. However, there was a high correlation
between the duration of increased glutamate levels (from start to drop) and the duration of REM sleep
(r=0.92, df 45, p<0.0001; Fig. 3B). The time taken for glutamate levels to fall from peak to basal
values (22.4 ± 4.0 sec) with REM sleep offset was similar to the time taken for glutamate levels to
increase from start to peak values (27.7 ± 4.2 sec) with REM sleep onset (Table 1A).
Variation in glutamate levels was calculated relative to AW onset and offset (Fig.5). We observed that
glutamate levels started to increase after the onset of AW (5.4 ± 15.7 sec) but started to decrease prior
to the offset of AW (47.2 ± 25.6 sec). Glutamate levels increased in AW by 0.32 ± 0.05 μM compared
to the preceding state. The magnitude of the relative changes in glutamate levels with AW was similar
when AW was preceded by either NREM or QW. We did not observe any changes in glutamate levels
linked with muscle tone. There was a high correlation between the duration of AW and the duration of
increased glutamate levels (r=0.91, df 16, p<0.0001; Fig.6A). The magnitude of increase in glutamate
levels was also significantly correlated with the duration of AW (r=0.63, df 16, p<0.01, Fig. 6B). The
time taken for glutamate levels to reach peak values after onset of an AW period (140.3 ± 40.8 sec)
was nonsignificantly longer than the time taken for glutamate levels to return to basal values in the
subsequent state (89.5 ± 15.8 sec; Table 1B).
Short periods of sleep deprivation by gentle handling (20-120 min) produced gradual increases in
glutamate levels in the PH-TMN. Glutamate sensor data showed that glutamate levels peaked 15-20
min after the initiation of sleep deprivation (forced waking). Glutamate levels increased up to 1.51 μM
from basal values. Termination of sleep deprivation led to a gradual reduction in glutamate levels.
Spontaneous waking increased glutamate levels by 0.32 μM.
Glutamate levels in microdialysis samples collected during REM sleep were higher than those
collected during AW and NREM sleep (AW= 2.3 ± 0.12; NREM= 2.4 ± 0.06, REM= 2.5 ± 0.17
pmol/sample). However, these differences were not statistically significant. Microdialysis samples
from the PH-TMN also showed a nonsignificant increase in glutamate levels with sleep deprivation
compared to spontaneous waking (2.8 ± 0.02 vs. 2.5 ± 0.04 pmol/sample). These results are consistent
with results obtained with the glutamate biosensor, although they lack its temporal resolution (Fig.7).
Perifornical–lateral hypothalamus (PF-LH):
The highest glutamate levels in the PF-LH were observed during AW compared to NREM sleep (27
episodes from 3 rats, Fig.8A), as in the PH-TMN. There was a high correlation between AW duration
and the duration of increased glutamate levels (r=0.91, df 26, p<0.0001, Fig.8B). The time taken for
glutamate levels to decrease at the offset of AW was significantly longer than the time taken for it to
increase at the onset of AW (16.7 ± 3.8 sec versus 8.6 ± 1.3 sec, t= 2.0, df 54, p<0.05, Table 1B)
Glutamate levels increased in REM sleep (by 0.42 ± 0.09 μM) compared to the prior NREM sleep
periods. This was also observed in the PH-TMN. Sleep deprivation produced a non-significant
increase in glutamate levels compared to spontaneous AW (0.50 ± 0.08 μM with sleep deprivation vs.
0.52 ± 0.05 μM in AW).
Glutamate levels in the frontal cortex increased with AW (16 episodes) and REM sleep (16 episodes)
compared to NREM sleep (from 3 rats, Fig.9A). Similar findings were observed in the PH-TMN and
PF-LH. There was a significant correlation between AW duration and the duration of increased
glutamate levels (r=0.67, df 15, p<0.01, Fig.9B). Sleep deprivation increased glutamate levels
significantly more than spontaneous AW (0.72 ±0.14 μM with sleep deprivation vs. 0.35 ±0.05 μM in
AW; t=3.0, df 21, p<0.007). REM sleep duration was also significantly correlated with the duration of
increased glutamate levels (r=0.65, df 15, p<0.01). The time taken for glutamate levels to increase
with AW was significantly longer than the time taken for glutamate levels to decrease to basal levels
(120.6 ± 21.0 sec. vs. 68.1 ± 12.3 sec; t=2.2, df 30, p<0.03; Table 1B). This was also the case in REM
sleep (42.4 ± 6.6 sec. vs. 20.3 ± 3.5 sec; t=3.0, df 32, p<0.005; Table 1A).
The time taken for glutamate levels to increase, from ‘start to peak’, and to decrease, from ‘drop to
baseline’, was faster in the PF-LH compared to both the PH-TMN and the frontal cortex. Furthermore,
glutamate levels increased at a faster rate than it decreased, during REM/AW onset and offset
respectively, in the PF-LH. This is in contrast to the faster decrease in glutamate levels compared to its
increase in the PH-TMN and frontal cortex (Table 1).
Analysis of variance of the changes in glutamate levels and of the time course of glutamate variation at
REM sleep onset (20 sec prior to and 20 sec after REM onset in 5 sec periods) showed a significant
difference in glutamate levels across the three brain regions (F (2, 92) =9.2, P<0.0001, ANOVA).
There was also a significant effect of time course on glutamate levels (F (7, 322)=40.8, P<0.0001,
ANOVA) and the interaction of these two variables (F (14, 644)=9.6, P<0.0001, ANOVA). On the
other hand, at REM sleep offset (30 sec prior to and 15 sec after the REM offset), glutamate levels did
not significantly differ across the three brain regions. The effect of time course on glutamate levels (F
(8, 368)=30.3, P<0.0001, ANOVA) and the interaction of these two variables (F (16, 736)=7.7,
P<0.0001, ANOVA) however, were significant.
Analysis of variance of the changes in glutamate levels at AW onset showed there was no significant
difference in glutamate levels (20 sec prior to and 20 sec after the AW onset) across the different brain
regions. However, there were significant effects of time on glutamate levels (F (7, 231)=7.2,
15 Download full-text
P<0.0001, ANOVA) and a significant interaction of these two variables (F (14, 462)=2.8, P<0.0006,
ANOVA). In contrast, AW offset (30 sec prior to and 15 sec after the AW offset) showed no
significant difference in glutamate levels across brain regions. However, there were significant effects
of time on glutamate levels (F (8, 256)=10.2, P<0.0001, ANOVA) and a signifcant interaction of these
two variables (F (16, 512)=1.6, P<0.05, ANOVA).
We observed increased glutamate levels with AW and REM sleep compared to NREM sleep. The
increase in glutamate levels strongly coincided with REM sleep onset (0.6±1.8 sec) and less strongly
with AW onset (5.4 ± 15.7 sec) in the PH-TMN. Decreased glutamate levels preceded the offset of
AW and REM sleep. These findings suggest that processes linked to reduced glutamate levels may be
involved in the termination of AW and REM sleep states.
Glutamate is not enzymatically broken down in the synaptic cleft, however, evidence suggests that
most free glutamate is cleared from the cleft very rapidly (10, 47), perhaps within 1 msec (6). In
theory, such rapid clearance might be accomplished by diffusion alone (49). However, a study by
Tong & Jahr (45) suggested that binding to glutamate transporters speeds the clearance of free
glutamate. This rapid turnover of glutamate may not be detectable by the standard microdialysis-
HPLC techniques, which require much longer duration samples.
Simultaneous glutamate measurements by microdialysis and by the glutamate biosensor were recorded
in the PH-TMN. Consistent with the findings of Nitz and Siegel (28), we did not observe any
significant variation in glutamate levels across sleep-wake states using the microdialysis technique.