Adenosinergic modulation of rat basal forebrain neurons during sleep and waking: neuronal recording with microdialysis

Abstract

1. The cholinergic system of the basal forebrain (BF) is hypothesized to play an important role in behavioural and electrocortical arousal. Adenosine has been proposed as a sleep-promoting substance that induces sleep by inhibiting cholinergic neurons of the BF and brainstem. However, adenosinergic influences on the activity of BF neurons in naturally awake and sleeping animals have not been demonstrated.

2. We recorded the sleep-wake discharge profile of BF neurons and simultaneously assessed adenosinergic influences on wake- and sleep-related activity of these neurons by delivering adenosinergic agents adjacent to the recorded neurons with a microdialysis probe. Discharge rates of BF neurons were recorded through two to three sleep-wake episodes during baseline (artificial cerebrospinal fluid perfusion), and after delivering an adenosine transport inhibitor (s-(p-nitrobenzyl)-6-thioinosine; NBTI), or exogenous adenosine, or a selective adenosine A1 receptor antagonist (8-cyclopentyl-1,3-dimethylxanthine; CPDX).

3. NBTI and adenosine decreased the discharge rate of BF neurons during both waking and non-rapid eye movement (NREM) sleep. In contrast, CPDX increased the discharge rate of BF neurons during both waking and NREM sleep. These results suggest that in naturally awake and sleeping animals, adenosine exerts tonic inhibitory influences on BF neurons, supporting the hypothesized role of adenosine in sleep regulation.

4. However, in the presence of exogenous adenosine, NBTI or CPDX, BF neurons retained their wake- and sleep-related discharge patterns, i.e. still exhibited changes in discharge rate during transitions between waking and NREM sleep. This suggests that other neurotransmitters/neuromodulators also contribute to the sleep-wake discharge modulation of BF neurons.

The basal forebrain (BF), including the nuclei of the vertical and horizontal limbs of the diagonal band of Broca, magnocellular preoptic area, ventral pallidum, substantia innominata and nucleus basalis, contains a system of cholinergic neurons that project to the entire neocortex and limbic telencephalon (Mesulam et al. 1983; Saper, 1984; Woolf et al. 1984). BF cholinergic neurons exert excitatory effects on cortical targets (Rasmusson & Dykes, 1988; Hars et al. 1993) and play an important role in EEG desynchronization and behavioural arousal (for review see Semba, 1991; Jones, 1993; Szymusiak, 1995). The spontaneous release of cortical acetylcholine (ACh) is higher during waking than during non-rapid eye movement (NREM) sleep (Phillis, 1968; Jasper & Tessier, 1971). Electrical stimulation of the BF increases cortical ACh release in anaesthetized rats (Kurosawa et al. 1989; Rasmusson et al. 1992) and evokes EEG desynchronization (Casamenti et al. 1986). Neurotoxic lesion of BF cholinergic neurons decreases cortical ACh release (Dekker & Thal, 1993) and increases EEG slow wave activity (Stewart et al. 1984; Buzsaki et al. 1988; Riekkinen et al. 1990; Bassant et al. 1995). Extracellular neuronal recording studies in the BF of cats (Detari et al. 1984; Szymusiak & McGinty, 1986, 1989) and rats (Koyama & Hayaishi, 1994; Alam et al. 1997) have identified wake-related neurons to be a common BF cell type, though sleep-related and state-indifferent neurons have also been described.

Adenosine is an inhibitory neuromodulator (e.g. Phillis & Wu, 1981; for review see Dunwiddie, 1985; Greene & Haas, 1991; Chagoya de Sanchez, 1995) and is hypothesized to be involved in the regulation of sleep and arousal (Radulovacki et al. 1984; Rainnie et al. 1994; Benington & Heller, 1995). Adenosine or its agonists promote sleep and increase EEG slow wave activity, whereas adenosine antagonists, such as caffeine and theophylline, are potent behavioural stimulants and suppress sleep (Radulovacki et al. 1984; Virus et al. 1990; Benington et al. 1995; Landolt et al. 1995; Schwierin et al. 1996; for review see Fredholm et al. 1999). The neuronal production of adenosine is coupled with metabolic activity, which is higher during waking than during sleep (Pull & McIlwain, 1972; Maquet, 1995). In freely moving rats, the extracellular levels of adenosine are higher in neostriatum and hippocampus during the lights off period, i.e. the rat's active period (Huston et al. 1996). In cats the extracellular levels of adenosine in the BF are higher during waking as compared with sleep, exhibit a progressive increase with forced waking, and decline during subsequent sleep (Porkka-Heiskanen et al. 1997).

Administration by microdialysis of adenosine (Portas et al. 1997) or its transport inhibitor (Porkka-Heiskanen et al. 1997) into cholinergic regions of the BF decreases waking in cats. Neurons in the BF and cholinergic neurons in the mesopontine tegmentum are inhibited by adenosine in in vitro preparations (Rainnie et al. 1994). Since the cholinergic system of the BF and mesopontine tegmentum have been implicated in EEG desynchronization and behavioural arousal, and the neurons in these areas are inhibited by adenosine in vitro, it has been hypothesized that adenosine promotes sleep by inhibiting cholinergic neurons of the BF and brainstem (Rainnie et al. 1994). However, the influence of adenosine or its analogues on the activity of BF or mesopontine tegmentum neurons in freely behaving animals is unknown.

In this study, we recorded the sleep-wake discharge profile of BF neurons in freely behaving rats and simultaneously assessed adenosinergic influences on spontaneous wake- and sleep-related activity of these neurons. This was achieved by delivering either an adenosine transport inhibitor or exogenous adenosine, and an A1 adenosine receptor antagonist adjacent to the recorded neurons with a microdialysis probe.

METHODS

Experiments were performed on seven unanaesthetized and unrestrained male Sprague-Dawley rats, weighing between 250 and 350 g. These rats were maintained on a 12 h-12 h light-dark cycle (lights on from 08.00 h to 20.00 h), with food and water ad libitum. All experiments were conducted in accordance with the National Research Council Guide for the Care and Use of Laboratory Animals. Under surgical anaesthesia (ketamine + xylazine, 80:10 mg kg⁻¹i.p.) and aseptic conditions, rats were stereotaxically prepared for chronic recording of sleep-wake variables, BF neuronal activity, and microdialysis adjacent to the recording site. Electroencephalogram (EEG) and electromyogram (EMG) electrodes were implanted, according to standard techniques, for polygraphic determination of sleep-waking state. A single barrel (23 gauge stainless steel tube) mechanical microdrive was implanted such that its tip rested 3 mm above the dorsal aspect of the BF (A, −0.3 to −0.5; L, 2.00 to 2.5; H, 4.5 to 5; Paxinos & Watson, 1997). Five pairs of microwires, each consisting of two 20 μm insulated stainless steel wires glued together except for 2.0 mm at the tip, were passed through the barrel such that their tips projected into the BF. A microdialysis guide cannula (20 gauge stainless steel tubing) was implanted 0.6-0.7 mm lateral (centre to centre) to the microdrive and blocked with a stylet. A plastic syringe tube was fixed around the microdrive and guide cannula for their protection. Buprenorphine (0.03-0.1 mg kg⁻¹, once daily for 2 days) was given subcutaneously for 2 days for post-operative pain management.

Data acquisition

Experiments were begun 4–5 days after surgery. At least 12 h before the recording session, the stylet of the microdialysis guide cannula was replaced by a microdialysis probe with a semi-permeable membrane tip length of 1 mm, an outer diameter of 0.22 mm, and a molecular cut-off size of 50 kDa. The in vitro recovery rate of a group of randomly selected probes used in this study was 5–7 % at 37°C. The length of the probe was set such that the microwires were within 0.5-1.0 mm of the semi-permeable membrane and the estimated dialysis field of the probe included the extracellular environment of the recorded neurons. In this study, the microdialysis probe was fixed and microwires were advanced adjacent to the side of the exposed microdialysis membrane to minimize the tissue trauma and ensure maximum stability of the unit recording (Fig. 1A and B). The rat was placed in a sound-attenuated recording chamber (ambient temperature, 25 ± 1°C) and connected with cables for the recording of neurophysiological variables. The probe was continuously perfused with artificial cerebrospinal fluid (ACSF; composition (mm): 145 Na⁺, 2.7 K⁺, 1.0 Mg²⁺, 1.2 Ca²⁺, 1.5 Cl⁻ and 2 Na₂HPO₄, pH 7.2) at a flow of ∼2 μl min⁻¹ by a peristaltic pump kept outside the chamber. The time taken by the solution to reach from the reservoir to the tip of the probe was precisely calibrated.

Figure 1. Microdialysis with adjacent unit recording.

A, a schematic representation of the microdrive and microdialysis probe arrangements. The microwires were carried in a screw-driven microdrive, which was adjacent to the fixed microdialysis probe. The recorded neurons were within the perfusion area of the microdialysis probe (0.5-1.0 mm). B, photomicrograph of a brain section showing microwire tracts (small arrows) and the tract of the microdialysis probe (large arrows). ac, anterior commissure; 3V, third ventricle; oc, optic chiasm.

The microdrive was advanced in 25–30 μm steps until an isolated single unit with signal-to-noise ratio > 3.0, as confirmed in oscilloscopic traces, was found. EEG, EMG, raw microwire signals and window discriminator output were digitized (Cambridge Electronic Design 1401; supporting software, Spike2) and stored on a disk for subsequent analysis. Multiple spikes, if present, were sorted on the basis of spike shape parameters (spike amplitude, duration), from the amplified unprocessed microwire signals. EEG, EMG, window discriminator and spike integrator outputs were also recorded on a polygraph.

The discharge rate of an isolated BF neuron was recorded through two to three stable sleep-wake cycles (∼30–45 min) with ACSF perfusion as the control (baseline). After baseline recording, drugs of known concentration were dialysed into the extracellular environment of the recorded neurons for 10–15 min. This study was aimed at investigating the transient effects of adenosinergic agents on the activities of the local neuronal population without triggering a global behavioural response. Therefore, the drugs were dialysed for a short duration and doses of drugs were adjusted to detect responses of neurons without changing behavioural state. After delivery of drugs, the perfusion medium was switched back to ACSF and the recording continued for another 45–90 min as wash out or recovery. The drugs used in this study included: 1–10 μms-(p-nitrobenzyl)-6-thioinosine (NBTI; Sigma), an adenosine transport inhibitor; 300 μm adenosine (Sigma); and 1–5 μm 8-cyclopentyl-1,3-dimethylxanthine (CPDX; RBI), an adenosine A1 receptor antagonist. During the entire recording session, the animal was undisturbed unless it was necessary to achieve a stable episode of waking by mild auditory stimuli or gentle touch. Each neuron was tested once and only with a single dose of each drug.

Data analysis

Sleep-wake states were identified on the basis of EEG and EMG patterns using standard criteria (Timo-Iaria et al. 1970). Mean discharge rate per second during waking and NREM sleep was calculated from 40–200 s blocks (mean sample duration, ∼74 s during waking and ∼72 s during sleep) of stable waking and NREM sleep from two to three waking and NREM sleep episodes. Criteria for wake-related, state-indifferent and sleep-related neurons were adopted from previous publications (e.g. Alam et al. 1995a). The neurons were defined as wake related if the ratio of their NREM sleep discharge rate and wake discharge rate (NREM/wake discharge ratio) was < 0.8, i.e. firing rate during NREM sleep was at least 20 % lower than during waking. The neurons were classified as sleep related if the NREM/wake discharge ratio was > 1.2, i.e. their discharge rate was at least 20 % higher during NREM sleep than during waking. The neurons having a NREM/wake discharge ratio of > 0.8 and < 1.2 were considered as state-indifferent neurons. Since drugs were perfused for a short duration, and REM sleep was seldom observed during drug administration, only waking and NREM sleep were analysed. The Wilcoxon matched-pair signed rank test was used to determine the statistical significance of the drug treatments on neuronal activity during waking and NREM sleep.

Histology

At the end of the recording session, under deep anaesthesia (pentobarbital, 100 mg kg⁻¹i.p.), microlesions were made at the tip of one to two microwires from which most cells were recorded by passing a small current (20 μA, 15–20 s). The animals were then injected with heparin (500 U i.p.), and perfused transcardially with 30–50 ml of 0.1 M phosphate buffer, pH 7.4, then 300 ml of 4 % paraformaldehyde and 100 ml of 10 and 30 % sucrose in phosphate buffer. The brains were removed and equilibrated in 30 % sucrose and then freeze-sectioned at 40 μm thickness. Immunohistochemistry for choline acetyltransferase (ChAT), a marker of cholinergic neurons, was performed on a series containing every third section. Sections were rinsed in Tris-buffered saline (TBS) and incubated in blocking solution consisting of 0.2 % Triton X-100 and 4 % normal goat serum in TBS, pH 7.0, for 2 h at 4°C. The sections were then rinsed and incubated in primary antisera, consisting of mouse anti-ChAT (Boehringer Mannheim Biochemica, Germany) diluted 1:1000 in diluent (0.1 % Triton X-100 and 4 % normal goat serum in TBS) for 48 h at 4°C. The sections were then rinsed in TBS and incubated in diluent containing goat anti-mouse IgG (1:200; Jackson Labs) for 2 h at 4°C, and then rinsed and incubated in mouse peroxidase-anti-perioxidase (1:200; Sternberger Monoclonals, MD, USA) for 2 h at room temperature. Sections were then rinsed in 0.05 M Tris buffer, reacted for 5–10 min in 0.05 % diaminobenzidine (Sigma) and 0.001 % hydrogen peroxide in 0.05 M Tris buffer, and rinsed in TBS. Following immunohistochemical procedures, sections were mounted onto gelatin-alum-coated glass microscope slides, dehydrated and covered with glass coverslips and DPX mountant. The locations of the microdialysis probe and the microwire tracts were histologically identified in Nissl-stained sections and the positions of the recorded neurons were plotted (Fig. 8). The distribution of the ChAT-positive neurons in the recorded area was mapped.

Figure 8. Line drawings of two representative brain sections showing the areas of neuronal recording at two rostral-caudal planes.

Left side of the sections shows the distribution of the neurons studied for the effects of 10 μm NBTI, 300 μm adenosine and 5 μm CPDX, according to their sleep-wake discharge patterns. ChAT-positive neurons are plotted on the right side of the section (•). Most of the recorded neurons were within or close to areas containing ChAT-positive neurons. ac, anterior commissure; oc, optic chiasm; 3V, third ventricle.

RESULTS

The spontaneous neuronal activity of 85 BF neurons was recorded across the sleep-wake cycle during microdialysis perfusion of ACSF (control) and during perfusion of NBTI, adenosine or CPDX. A typical experiment is shown in Fig. 2. Waking and NREM sleep samples were obtained during baseline ACSF perfusion, drug perfusion and recovery. During ACSF perfusion neuronal recordings exhibited a stable signal-to-noise ratio and the consistent wake-related, state-indifferent or sleep-related discharge patterns typical of BF neurons as described previously (Szymusiak & McGinty, 1986, 1989; Alam et al. 1995b, 1997). The microdialysis perfusion of drugs changed neuronal discharge rate in a specific and dose-dependent manner without affecting spike shape parameters. Changes in discharge rate could be seen within < 5 min of the arrival of the drug at the microdialysis membrane and were maintained throughout the treatment. Neither ACSF nor drug perfusion induced any clear changes in the behavioural states of the animals. During recovery, discharge rates returned to baseline levels, although drug effects generally outlasted the duration of drug perfusion by 5–30 min.

Figure 2. Effects of microdialysis perfusion of 10 μm NBTI on the neuronal activity of an individual wake-related neuron during waking and NREM sleep.

A, a 48 min continuous recording session showing the discharge rate (spikes s⁻¹) of a neuron during baseline, NBTI perfusion and recovery. The discharge rate of this neuron increased during each arousal state indicated by elevation of EMG activity and desynchronized EEG. This neuron also exhibited an increased discharge during REM sleep, as is typical of BF neurons. NBTI was perfused for about 12 min. A strong suppression of the neuronal activity after NBTI delivery, and recovery following wash out can be seen. Lower panels are 5 min expanded tracings of the areas marked in A showing wake- and sleep-related neuronal discharge rate during baseline (B), NBTI perfusion (C) and recovery (D). In response to NBTI, this neuron exhibited a strong suppression of discharge during waking (C). During NREM sleep, this neuron had a low baseline discharge rate but exhibited further suppression of discharge due to NBTI perfusion. The action potentials to the right of B–D represent the average waveforms of all the action potentials recorded during the representative section in each condition. There was no change in spike shape or amplitude as a result of microdialysis perfusion of ACSF or NBTI. EEG, electroencephalogram; EMG, electromyogram; Unit, neuronal activity.

In five neurons, doses of 1–5 μm NBTI (two at 1 μm and three at 5 μm) produced marginal changes in discharge rate during waking (0.80 ± 0.20 vs. 0.72 ± 0.27 spikes s⁻¹; means ±s.e.m.) and NREM sleep (0.43 ± 0.12 vs. 0.32 ± 0.10 spikes s⁻¹). A concentration of 10 μm NBTI was consistently effective and was used for the rest of the neurons (n= 35). We also examined the effects of exogenous adenosine on BF neurons to compare them with the effects produced by NBTI. It was shown in an earlier study that microdialysis perfusion of 300 μm adenosine into the BF of cats produce a decrease in waking (Portas et al. 1997). Therefore, the effects of 300 μm adenosine on neuronal activity were examined. A dose of 1 μm CPDX produced marginal changes in discharge rate (n= 3) during waking (4.64 ± 1.53 vs. 4.78 ± 1.65 spikes s⁻¹) as well as during NREM sleep (0.70 ± 0.46 vs. 0.78 ± 0.29 spikes s⁻¹). A dose of 5 μm was consistently effective and was used for the rest of the neurons (n= 31). The details of the results are as follows.

Wake-related neurons

The effects of NBTI on the discharge rate of an individual wake-related neuron and on the mean discharge rate of wake-related neurons during waking and NREM sleep are shown in Fig 2 and Fig 3, respectively. Of the neurons studied with NBTI, the majority were wake related (n= 20, 57 %). These neurons had a mean NREM/wake discharge ratio of 0.47 ± 0.05, i.e. their mean discharge rate during NREM sleep was 53 % lower than their discharge rate during waking. The discharge rate of wake-related neurons was significantly reduced by NBTI during both waking (-50 ± 5 %; range, −98 to 5 %; Z=−3.742, two-tailed P= 0.0002, Wlicoxon matched-pair signed-rank test, throughout) and NREM sleep (-41 ± 8 %; range, −91 to 42 %; Z=−3.299, P= 0.001; Fig 2 and Fig 3A and B). The NBTI-induced decrease in mean discharge rate during waking (baseline discharge rate vs. discharge rate during NBTI perfusion) was not significantly different from the decrease during NREM sleep. Therefore, the NREM/wake discharge ratio of wake-related neurons after NBTI treatment was not significantly different from the baseline NREM/wake discharge ratio (Z=−1.176, P= 0.239; Fig. 3A and B). The effects of adenosine on the discharge rate of an individual wake-related neuron, and on the mean discharge rate of wake-related neurons during waking and NREM sleep are shown in Fig 4 (Unit 1) and Fig 5, respectively. Adenosine perfusion also decreased the activity of wake-related neurons (n= 5) during waking (-33 ± 8 %, Z=−2.022, P= 0.0431) as well as during NREM sleep (-28 ± 14 %, Z=−2.022, P= 0.0431) without changing the NREM/wake discharge ratio (Z=−0.944, P= 0.345; Fig. 5A and B).

Figure 3. Neuronal responses to NBTI.

The effects of 10 μm NBTI on discharge rate (spikes s⁻¹; means and s.e.m.) of wake-related neurons (A), state-indifferent neurons (C) and sleep-related neurons (E) during waking and NREM sleep. The NREM/wake discharge ratios of wake-related (B), state-indifferent (D) and sleep-related (F) neurons during baseline and after NBTI perfusion are also shown. NBTI suppressed the discharge rate of wake-related and state-indifferent neurons during both waking and NREM sleep without producing any significant change in the NREM/wake discharge ratio. Sleep-related neurons showed suppression only during waking resulting in an increased NREM/wake discharge ratio. , control (ACSF perfusion); ▪, 10 μm NBTI perfusion; * P < 0.05, ** P < 0.01 (Wilcoxon matched-pair signed rank test).

Figure 4. Effects of microdialysis perfusion of 300 μm adenosine on the discharge rate of simultaneously recorded wake-related (Unit 1) and sleep-related (Unit 2) neurons during waking and NREM sleep.

Neuronal discharge is shown during baseline ACSF (A) and adenosine perfusion (B). The action potentials to the right of A and B represent the average waveforms of all the action potentials recorded for each neuron during the representative sections. Although the action potentials from the two neurons were similar in shape, they differed in terms of their spike amplitude and sleep-wake discharge patterns. Adenosine suppressed the mean discharge rate of this wake-related neuron during waking and NREM sleep. In contrast, adenosine increased the discharge rate of this sleep-related neuron during waking and NREM sleep. Abbreviations as in Fig. 2.

Figure 5. Neuronal responses to adenosine.

The effects of 300 μm adenosine on discharge rate (spikes s⁻¹; means and s.e.m.) of wake-related neurons (A), state-indifferent neurons (C) and sleep-related neurons (E) during waking and NREM sleep. The NREM/wake discharge ratios of wake-related (B), state-indifferent (D) and sleep-related (F) neurons during baseline and after adenosine perfusion are also shown. Note that adenosine produced effects similar to those produced by NBTI. , control (ACSF perfusion); ▪, 300 μm adenosine perfusion; * P < 0.05 (Wilcoxon matched-pair signed rank test).

The effects of CPDX on the discharge rate of an individual wake-related neuron and on the wake-related neuronal population during waking and NREM sleep are shown in Fig 6A and B, and Fig 7, respectively. The majority of neurons studied with CPDX were wake related (n= 18, 58 %). These neurons had a baseline NREM/wake discharge ratio of 0.46 ± 0.04, i.e. their discharge rate during NREM sleep was 54 % lower than their discharge rate during waking. In response to CPDX, wake-related neurons exhibited a significant increase in discharge during waking (126 ± 48 %; range, −38 to 667 %; Z=−2.896, P= 0.004) as well as during NREM sleep (72 ± 20 %; range, −20 to 308 %; Z=−3.113, P= 0.002; Fig 6A and B, and Fig 7A and B). Since the increase in discharge rate during waking was not significantly higher than the increase during NREM sleep, the NREM/wake discharge ratio was not changed significantly (Z=−0.500, P= 0.617; Fig. 7B).

Figure 6. Effects of 5 μm CPDX on the neuronal activity of an individual wake-related neuron (A and B) and a sleep-related neuron (C and D) during waking and NREM sleep.

The action potentials to the right of A–D represent the average waveforms of all the action potentials recorded during the representative sections in each condition. CPDX increased the discharge rate of this wake-related neuron during both waking and NREM sleep (B). CPDX increased the discharge rate of this sleep-related neuron during waking with a smaller effect during NREM sleep (D). Abbreviations as in Fig. 2.

Figure 7. Neuronal responses to CPDX.

The effects of 5 μm CPDX on discharge rate (spikes s⁻¹; means and s.e.m.) of wake-related neurons (A), state-indifferent neurons (C) and sleep-related neurons (E) during waking and NREM sleep. The NREM/wake discharge ratios of wake-related (B), state-indifferent (D) and sleep-related (F) neurons during baseline and after CPDX perfusion are also shown. In response to CPDX, all neuronal types exhibited increased discharge during both waking and NREM sleep. However, sleep-related neurons exhibited a smaller increase in discharge during NREM sleep resulting in a significant decrease in their NREM/wake discharge ratio. , control (ACSF perfusion); ▪, 5 μm CPDX perfusion; * P < 0.05; ** P < 0.01 (Wilcoxon matched-pair signed rank test).

State-indifferent neurons

Similar to the effects observed in wake-related neurons, NBTI decreased the activity of state-indifferent neurons (n= 10) during both waking (-42 ± 10 %, Z=−2.803, P= 0.005) and NREM sleep (-42 ± 11 %, Z=−2.391, P= 0.012) without changing the NREM/wake discharge ratio (Z=−0.764, P= 0.444; Fig. 3C and D). In response to adenosine, state-indifferent neurons (n= 3) also exhibited a decrease in discharge rate during waking and NREM sleep (Fig. 5C and D).

CPDX significantly increased the discharge rate of state-indifferent neurons (n= 7) during waking (51 ± 24 %, Z=−2.197, P= 0.028) as well as during NREM sleep (53 ± 38 %, Z=−2.028, P= 0.042) without significantly affecting the NREM/wake discharge ratio (Z=−0.169, P= 0.863; Fig. 7C and D).

Sleep-related neurons

NBTI significantly reduced the activity of sleep-related neurons (n= 5) during waking (-28 ± 11 %; range, −6 to −70 %; Z=−2.022, P= 0.043) but had no effect on discharge rate during NREM sleep (4 ± 22 %; range, −58 to 74 %; Z=−0.404, P= 0.685; Fig. 3E and F). The mean NREM/wake discharge ratio of these neurons was 1.49 ± 0.09, i.e. their mean discharge during NREM sleep was 51 % higher than their discharge during waking. During NREM sleep, two of five neurons showed increased discharge whereas three exhibited decreased discharge in response to NBTI. NBTI significantly increased the NREM/wake discharge ratio of these neurons (Z=−2.022, P= 0.0431; Fig. 3E and F). In response to adenosine, of three sleep-related neurons, two exhibited a decreased discharge during waking and NREM sleep, whereas one neuron showed increased discharge during both waking and NREM sleep (Fig 4A and B (Unit 2), and Fig 5E and F).

CPDX significantly increased the discharge rate of sleep-related neurons (n= 6) during waking as well as during sleep (Fig 6C and D, and Fig 7E and F). These neurons had a mean NREM/wake discharge ratio of 1.59 ± 0.1, i.e. their mean discharge during NREM sleep was 41 % higher than their discharge during waking. The CPDX-induced increase in discharge during waking was higher than that during NREM sleep (104 ± 26 %; range, 35 to 190 %vs. 49 ± 27 %; range, −27 to 150 %; Z=−2.201, P= 0.027), significantly reducing the NREM/wake discharge ratio (Z=−2.201, P= 0.027; Fig. 7E and F).

Locations of the recorded neurons

Figure 8 shows the locations of the recorded neurons coded by their sleep-wake discharge profile. The cells were distributed in basal forebrain sites including areas with dense cholinergic neuronal populations, namely, the magnocellular preoptic area, substantia innominata and horizontal limb of the diagonal band (Fig. 8).

DISCUSSION

In this study, adenosinergic influences on the activity of BF neurons were examined in freely behaving rats. The extracellular neuronal activity of BF neurons was recorded with microdialysis perfusion of ACSF (baseline), an adenosine transport inhibitor, exogenous adenosine and an adenosine A1 receptor antagonist, adjacent to the recorded neurons. The resulting changes in discharge rates were neuromodulatory effects of the drugs and not artifactual as shown by the following observations.

The drug effects were consistent, reversible, dose dependent, and time linked to the period of drug administration. The continuous perfusion of ACSF alone did not produce any changes in the discharge rate. Adenosine or its transport inhibitor and the adenosine A1 receptor antagonist had opposite effects on neuronal activity, as predicted. Computer-generated templates, based upon spike shape parameters used for spike sorting, eliminated any possible artifactual spikes and ensured that activity of the same neuron was recorded across the baseline, drug delivery and recovery phases (see Figs 2, 4 and 6).

The strengths of this technique include the capability for stable unit recording during drug delivery, controlled timing of drug administration, the potential for studying the same neuron repeatedly for the same drug or both agonist and antagonist, and the capacity for coincident behavioural assessment. This procedure provides an alternative to microinjection or microiontophoresis and can be adopted as a reliable and artifact-free technique for behavioural and pharmacological characterization of neurons. A similar technique has been described previously (Ludvig et al. 1994; Thakkar et al. 1998).

Depending on the location of the microdialysis probe and the concentration of drug delivered, behavioural responses can be avoided or elicited (Porkka-Heiskanen et al. 1997; Portas et al. 1997; Thakkar et al. 1998) according to the preference of the experiment. The goal of this study was to examine the transient effects of adenosinergic agents on the activity of local BF neurons during spontaneously occurring sleep and waking. Therefore, low concentrations of adenosinergic agents were perfused unilaterally in the vicinity of BF neurons. Although in these experiments drug exposure was short and no remarkable changes in behavioural states were observed (see Fig. 2), long term administration of adenosinergic agents by microdialysis can alter sleep/waking ratios as shown previously (Porkka-Heiskanen et al. 1997; Portas et al. 1997).

The wake-related neurons studied were distributed in the magnocellular preoptic area, substantia innominata and horizontal limb of the diagonal band. These are the areas where cholinergic neurons are also present in abundance (see Fig. 8; see also Semba, 1991; Jones, 1993; Szymusiak, 1995). It is well documented that cholinergic neurons play an important role in behavioural arousal and are strongly state modulated (see Introduction). Our wake-related neuronal sample exhibited a sleep-wake discharge profile and phasic activation within waking similar to neurons which had been tentatively identified as cholinergic based on projection patterns and spike features (Szymusiak & McGinty, 1989). However, it is pertinent to mention that depending upon the particular BF region in rats, 20–50 % of cortically projecting neurons can be non-cholinergic (Rye et al. 1984; Wainer et al. 1985). Cholinergic neurons are typically large and therefore more likely to be resolved by extracellular electrodes. Although our method does not permit conclusive identification of neurotransmitter phenotypes of recorded cells, it is reasonable to hypothesize that at least some of the wake-related neurons in our sample were cholinergic.

This is the first study, to our knowledge, showing a tonic inhibitory influence of adenosinergic agents on BF neurons in naturally awake and sleeping animals. Our results show that NBTI consistently decreased neuronal activity of wake-related BF neurons. This inhibition of neuronal activity is hypothesized to be due to extracellular adenosine accumulated as a result of the transport inhibition. This interpretation is supported on the grounds that (i) adenosine dialysed into the BF in this study produced comparable results, and (ii) in earlier studies in freely moving cats, perfusion of NBTI was shown to produce increases in adenosine levels recovered in the same microdialysis probe (Porkka-Heiskanen et al. 1997). In contrast, the A1 adenosine receptor antagonist CPDX increased the discharge of BF neurons. The suppression of neuronal activity with adenosine or its transport blocker and an increase in activity with the adenosine A1 receptor antagonist is in agreement with an established role of adenosine as a primarily inhibitory neuromodulator in the central nervous system and the role of A1 receptors in this inhibition (e.g. Phillis & Wu, 1981; for review see Greene & Haas, 1991; Chagoya de Sanchez, 1995). These results are also consistent with a previous in vitro study in BF neurons (Rainnie et al. 1994).

Most wake-related neurons were strongly inhibited by adenosinergic agonists. This is consistent with a possible role of adenosine in promoting sleep by inhibiting wake-related neurons. However, state-indifferent neurons were also inhibited by adenosinergic agonists and facilitated by antagonists. Thus, the adenosinergic effects are not specific to state-related neurons. Moreover, in the presence of adenosine agonists and antagonist, wake-related neurons still displayed significant reductions in discharge rate during transitions from wakefulness to NREM sleep such that their NREM/wake discharge ratios did not change (see Figs 3, 5 and 7). This suggests that additional neurotransmitters/neuromodulators participate in regulating the sleep-wake state-dependent changes in activity of BF neurons.

While in vitro studies demonstrate the ability of adenosine to directly inhibit cholinergic neurons in the brainstem (Rainnie et al. 1994), the effects of adenosine analogues on BF neurons recorded in vivo, as described here, could have been the result of direct postsynaptic effects, presynaptic effects, or a combination of both. Adenosine is known to exert inhibitory modulatory effects on excitatory neurotransmitters including glutamate (Corradetti et al. 1984; Fastbom & Fredholm, 1985). Endogenous glutamate appears to play an important role in regulating the excitability of the BF cholinergic system as local BF perfusion with glutamate antagonists can prevent the increase in the neocortical ACh release evoked by brainstem stimulation (Rasmusson et al. 1994). Therefore, suppression of discharge in BF wake-related neurons evoked by adenosine or NBTI may have resulted, wholly or in part, from the presynaptic modulation of glutamate release.

The number of sleep-related neurons examined for adenosinergic influences was limited because we selected study sites with the highest density of cholinergic neurons. Although evidence is preliminary and responses were heterogeneous, some sleep-related neurons were unique in that they exhibited excitation in response to adenosine/NBTI and decreased discharge rate in response to CPDX during NREM sleep. As a group, sleep-related neurons exhibited an increase in NREM/wake discharge ratio in response to NBTI and a decrease in this ratio in response to CPDX. These responses are also consistent with the hypothesis of a sleep-promoting role for adenosine. Since A1 receptors are inhibitory in nature, it is possible that the increase in sleep-related neuronal discharge was due to the inhibition of inhibitory influences on these neurons during NREM sleep, perhaps originating on wake-related neurons. This interpretation is consistent with a recent report that presumed sleep-active ventrolateral preoptic neurons, studied with intracellular recordings in vitro, exhibit loss of IPSPs in response to adenosine (Morairty et al. 1998). Alternatively, it is possible that the increase in discharge in sleep-related neurons evoked by adenosine or NBTI was mediated by A2 adenosine receptors. Although the distribution of A2 receptors in BF is not well characterized (Reddington & Lee, 1991), it has been shown that A2a adenosine receptor stimulation in rostral BF produces a marked increase in NREM as well as REM sleep (Satoh et al. 1996).