INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

PERIPHERAL ADMINISTRATION of lipopolysaccharide (LPS), a component of gram-negative bacterial endotoxin, is a frequently used model for the study of various aspects of bacterial pathogenesis and septicemia. LPS administration elicits many of the physiological perturbations commonly associated with bacterial infections such as fever, cytokine production and release, and initiation of the acute phase response. LPS administration is also associated with alterations in sleep and the electroencephalogram (EEG) (18, 28), as are gram-negative bacterial infections (41, 42). However, the mechanisms by which peripheral somnogens alter EEG patterns and induce sleep are unknown at present. Available data suggest that the sleep alterations produced during infectious disease are likely to be initially triggered by peripheral cytokines. Proposed mechanisms for the central transmission of peripheral signals for altered somnolence include the passage of humoral messengers into the brain at the choroid plexus or circumventricular organs (4, 11, 24, 34) and direct activation of peripheral vagal afferents by somnogens (7, 12, 16, 28, 43). Although somnogenic cytokines and their receptors have been detected in brain sites considered to be involved in the modulation of sleep and/or arousal (6, 20, 26), little is known about the effects of these putative somnogens at the neuronal level. Learning more about these issues is critical for understanding how sleep may be produced secondary to infectious disease and for determining whether cytokines modulate sleep under normal conditions.

Several regions of the BF, including the substantia innominata (SI) and the nucleus basalis (NB), and of the hypothalamus, including the lateral preoptic area (LPOA), contain neurons that demonstrate vigilance-related alterations in neuronal firing rates or in c-fos expression (3, 5, 22, 31, 36, 37, 39). To evaluate the potential involvement of these sites in mediating the somnogenic effects of LPS administration, we surveyed their neuronal firing patterns in unanesthetized rats, comparing patterns observed during spontaneous sleep and wakefulness to those observed during the same vigilance states after the intraperitoneal administration of LPS.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sleep and temperature assessment. Eight male Sprague-Dawley rats (300-350 g) were surgically implanted with instrumentation to permit monitoring of the EEG, the electromyogram (EMG), locomotor activity, and core temperature. Aseptic techniques were used for all surgical procedures. Rats were anesthetized with a pentobarbital-chloral hydrate mixture. Four insulated stainless steel screws (Plastics One, Roanoke, VA) were placed in the skull in bilateral frontal and parietotemporal positions to serve as EEG electrodes and a ground reference. EMG electrodes (Plastics One) were placed subcutaneously overlying the nuchal muscles. All electrodes were inserted into a headstage pedestal that was secured to the skull with dental acrylic (10). During the same surgery, rats were also implanted with subcutaneous transmitters (Data Sciences, St. Paul, MN) to telemetrically quantify locomotor activity and core temperature. After surgery, rats were housed in individual cages in a sound-attenuated temperature-controlled chamber under a 12:12-h light-dark cycle (lights on at 8:30 AM) at 23 ± 1°C and were handled daily.

Prior to the initiation of data collection, rats were given 1-2 wk to recover from surgery and to acclimate to the recording conditions. To permit collection of EEG and EMG data, rats were tethered to a six-channel electrical commutator with a lightweight cable that permitted unrestricted movement and were acclimated to the tether for at least 3 days prior to the start of data collection. After the acclimation period, sleep was monitored for 3 h in untreated rats (8:30-11:30 AM). At 11:30 AM ("time 0"), rats were injected intraperitoneally with 1 ml/kg of sterile pyrogen-free saline, and recording continued during the next 24 h. At 11:30 AM on the following day, rats were injected intraperitoneally with LPS (1 mg/kg; Escherichia coli serotype O26:B6, Sigma Chemical, St. Louis, MO), and recording continued until 8:30 AM the following morning. Injections were performed at 11:30 AM to correspond to the circadian conditions used in the neuronal recording studies described below.

EEG and EMG signals were processed through an eight-channel Grass polygraph. The EEG signals were passed through delta (1-4 Hz) and theta (4-8 Hz) filters (Coulbourn Instruments, LeHigh Valley, PA) and into a data-acquisition system (Cambridge Electronics Design, Cambridge, England) that samples, digitizes, and stores signals at 10 Hz. EMG signals were similarly processed without filtering. All data were continuously sampled and stored on a computer. Computer-assisted EEG and EMG analyses applying a customized software algorithm were used to assign a vigilance state to each 10-s epoch during the recording period. Initially, EEG tracings were visually examined to determine a threshold delta-wave amplitude (DWA) associated with slow-wave sleep (SWS) for each animal. Thresholds for EMG associated with periods of movement and for ratios of theta- to delta-band amplitudes associated with rapid eye movement sleep (REMS) were also determined. On the basis of these thresholds, the data for each animal were then scored by computer in 10-sec intervals for the entire experiment. An animal was considered to be in a state of SWS whenever the average DWA for any two consecutive intervals exceeded the SWS threshold in association with a low-amplitude EMG signal. REMS was identified by low-amplitude DWA and EMG signals that occurred in association with a high ratio of theta to delta amplitudes. At all other times, the animal was considered to be awake. All computer-scored data were visually reviewed to verify the accuracy of the computerized scoring. Data were summarized in 1-h intervals as the percentage of time spent in SWS and REMS and the average DWA during SWS.

Body temperature and locomotor activity were measured telemetrically via signals that were emitted from the intraperitoneal transmitters and detected by receivers (Data Sciences, St. Paul, MN) positioned under the cages. Temperatures were sampled every 10 min. Activity counts were summed across 10-min intervals. These data were stored on computer using a Dataquest III data-acquisition system (Data Sciences). Core temperature and locomotor activity data were summarized in 1-h intervals.

Neuronal recording. Thirty male Sprague-Dawley rats (300-420 g) were used for neuronal recording. Rats were anesthetized with a pentobarbital-chloral hydrate mixture, and aseptic technique was used to permanently implant multiwire electrodes (Tallent Technologies, Needham, MA) that were directed at various BF or hypothalamic targets (10). Electrodes consisted of sixteen 25-µm stainless steel wires attached to a microdrive that could be advanced in 0.1-mm increments. The electrode assembly was secured to the skull with acrylic cement. After surgery, rats were housed in individual cages under a 12:12-h light-dark cycle at 23 ± 1°C in sound-attenuated chambers. Recording sessions began no sooner than 1 wk after the surgery and were conducted during the light phase of the circadian cycle.

After the recovery period, the multiwire electrode assembly was advanced through the brain in increments of 0.1-0.2 mm/day. Electrodes were evaluated daily for the presence of discrete neuronal units with minimal signal-to-noise (S-N) ratios of 3:1. S-N ratios of recorded neurons typically ranged from 6:1 to 15:1. If acceptable units were simultaneously detected on multiple recording channels, unit activity was measured during epochs of both non-REMS (generally equivalent to EEG-defined SWS) and quiet wakefulness. Two criteria defined acceptable recording epochs. First, rats were required to maintain a stable behavioral state for a minimum of 2 min, and second, a minimum of 2,000 spikes were recorded during the epoch. If arousal from sleep or overt movement occurred before these criteria were met, data acquisition for that epoch was terminated, and the data were discarded. Recording was initiated again if the animal entered an appropriate behavioral state. Vigilance states were identified based on visual evaluation of the rat via a closed-circuit television camera located in the recording chamber. Non-REMS was defined as a state in which the rat was lying motionless in the cage, usually in a curled posture, with eyes closed and without the phasic movements typically associated with REMS. Quiet wakefulness was defined as a state in which the animal was alert but relatively inactive and not engaging in overt behaviors such as walking, rearing, grooming, eating, or drinking. Data collected in the sleep and temperature assessment described above revealed average non-REMS or SWS bout lengths of 1-2 min and ~10-20 bouts of non-REMS or SWS per hour, interspersed with epochs of REMS and/or wakefulness, during the light phase of the circadian cycle. These values are in close agreement with other reports in the literature (16).

Immediately after the initial recordings were completed, rats were injected intraperitoneally with LPS (1.0 mg/kg of Escherichia coli serotype O26:B6). Neuronal activity was then sampled during epochs of sleep and quiet wakefulness that occurred during the 60- to 100- and 150- to 210-min intervals after injection. Adherence to the epoch criteria described above, coupled with the need to wait until the rat spontaneously entered an appropriate vigilance state, necessitated the relatively broad postinjection intervals. Individual neurons were recorded continuously throughout the pre- and postinjection period and were monitored closely for stability of the waveform across time. Sample recordings obtained across time from four representative neurons are illustrated in Fig. 1.

View larger version (26K): [in this window] [in a new window]   Fig. 1.   Waveforms of representative neurons recorded across time. Panels show waveforms of individual neurons that were recorded continuously throughout pre- and postinjection period. Dual lines shown in each panel represent average waveform obtained from ~200 spikes superimposed on waveform from last spike comprising average. Signal-to-noise ratios ranged from 6:1 to 15:1. BF30-12, a nonvigilance-related ventral pallidal neuron with a lipopolysaccharide (LPS)-related decrease in firing rate; BF41-09, a wake-active parastrial neuron with an LPS-related increase in firing rate; BF51-03, a nonvigilance-related nucleus basalis neuron with an LPS-related decrease in firing rate; BF56-03, a nonvigilance-related substantia innominata neuron with an LPS-related increase in firing rate.

Rats were used for a maximum of three recording sessions in which they were injected with LPS. In all cases, an interval of at least 1 wk intervened between injection of LPS and resumption of electrode advancement for individual rats, resulting in a typical interval of at least 2 wk between repeated LPS injections. Of the 191 neurons recorded, 95 were recorded after the first LPS injection, 54 after the second injection, and 42 after the third. Similar results were obtained within each subgroup. For seven of the recorded neurons, the rat failed to enter an acceptable state of quiet wakefulness (i.e., an epoch adequate for the collection of at least 2,000 spikes over a minimum period of 2 min without extraneous movement or activity) during an extensive preinjection evaluation period, and, therefore, vigilance responsiveness could not be assessed for those neurons.

After the final recording session, rats were anesthetized and an electrolytic lesion (10 µA, 10 s) was made at the most ventral electrode penetration. The electrode was then gradually withdrawn, and additional marking lesions were made incrementally along the electrode tract. Two days after the lesions were made, rats were deeply anesthetized and transcardially perfused with saline followed by 10% Formalin. Coronal 40-µm sections were made on a cryostat and were stained for Nissl using standard techniques. The electrode tract and the location of lesions were evaluated microscopically to localize the recording sites (e.g., Fig. 2). The multiwire electrode was ~0.5 mm in diameter, and individual wires extended beyond the supporting cannula tip by 2.5-3.0 mm. Because the exact position of individual electrode tips could not be precisely determined, the histological identification of recording sites represents an approximation based on the location of the electrode used to make the lesion and the calculated depth of the recording sites ventral to the dura on each day of recording.

View larger version (142K): [in this window] [in a new window]   Fig. 2.   Example of histological identification of recording sites. Electrode tract and 2 electrolytic lesions are visible in this Nissl-stained coronal section. Locations of lesions and tract indicate that electrode passed through ventral pallidum/substantia innominata.

The responses of individual neurons were analyzed offline using spike-discrimination software (DataWave Systems, Thornton, CO). Custom software was used to determine the neuronal firing rates, interspike intervals, and autocorrelations. The ratio of baseline firing rates during sleep and quiet wakefulness and the ratio of firing rates during pre- and post-LPS epochs of sleep and wakefulness were calculated. On the basis of their firing properties, neurons were classified as vigilance related (sleep active or wake active) and nonvigilance related and as LPS responsive or nonresponsive. Vigilance-related neurons were operationally defined as those demonstrating a minimum change in firing rate of 50% during quiet wakefulness vs. non-REMS. Neurons were classified as sleep active if the sleep-to-wake firing rate ratio was 1.5 or greater and as wake active if the ratio was 0.5 or less. LPS-responsive neurons were defined as those demonstrating a minimum change in firing rate of 50% after LPS administration compared with the pre-LPS rate during the same vigilance state.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sleep and temperature assessment. Rats injected with saline showed clear circadian variation in the percentage of time spent in SWS, in DWA during SWS, and in core temperature (Fig. 3). After the administration of LPS, rats developed fevers of ~1°C that were present during postinjection hours 3-9. The percentage of time in SWS and DWA during SWS both increased modestly after 4-6 h; these effects persisted for 2-4 h. Locomotor activity and the percentage of time in REMS were not markedly altered after LPS injection (data not shown).

View larger version (27K): [in this window] [in a new window]   Fig. 3.   Sleep and temperature responses of LPS-treated rats. At "time 0" (11:30 AM, designated by vertical dashed line) on sequential days, rats (n = 8) were injected intraperitoneally with sterile pyrogen-free saline (1 ml/kg; day 1) or LPS (1 mg/kg; day 2). Panels represent percentage of time spent in slow-wave sleep (SWS) during each hour of recording (A), average delta-wave amplitude (DWA) during SWS (B; expressed as percentage of average value measured on baseline day of recording), and average hourly core temperature (C). Individual data points represent means ± SE. Dark bar on abscissa in B denotes dark phase of circadian cycle; lights were on from 8:30 AM to 8:30 PM.

Neuronal recording. Data were collected from 191 neurons in 30 rats before and after LPS administration. Brain structures surveyed included SI, NB, ventral globus pallidus, lateral hypothalamus (LH), ventrolateral caudate-putamen, LPOA, and parastrial nucleus (PS) (Table 1, Fig. 4). One-hundred seventeen units did not show LPS-related changes in firing rates, but 74 units (39% of the total) were responsive to LPS treatment (Table 1). Thirty-four of 184 units that were assessed during both sleep and quiet wakefulness before the administration of LPS showed vigilance-related changes in firing rates; 12 of these (7%) were sleep active, and 22 (12%) were wake active (Table 1). Twenty-seven of the LPS-responsive neurons (37%) also demonstrated vigilance-related responses (9 were sleep active and 18 were wake active; Table 2). In contrast, only seven of 111 non-LPS-responsive neurons (6%) demonstrated vigilance-related changes in firing rates (Table 2). Thus 79% of the vigilance-related neurons were LPS responsive compared with only 31% of non-vigilance-related neurons. Statistical analysis indicates a significant interaction between these two characteristics (² = 25.5, 1 df, P ± 0.001). However, the directionality of the LPS-related changes in firing rates was not significantly related to whether the neuron was sleep active or wake active (Table 3).

View larger version (21K): [in this window] [in a new window]   Fig. 4.   Location of LPS-responsive neurons. , Location of units that increased their firing rates after LPS administration; , location of neurons that fired more slowly after LPS treatment. Anterior-posterior levels 0.3 and 0.4 illustrate recording sites identified as lateral preoptic area (LPOA)/parastrial nucleus (PS), levels 0.8 and 0.92 illustrate sites identified as substantia innominata (SI)/nucleus basalis (NB), and level 1.3 illustrates sites identified as ventral globus pallidus (GP) and lateral hypothalamus (LH). Panels are modified from 29.

The primary brain structures surveyed were subregions of BF or hypothalamus. Most of these regions contained two populations of LPS-responsive neurons: those that increased their firing rates after LPS administration and those that decreased their firing rates (Fig. 4). The magnitude of the decreases in firing rates were similar across recording sites, averaging 33 ± 3% of the pre-LPS rates for units in SI/NB (n = 15) and 32 ± 6% for units in LPOA/PS (n = 6; Fig. 5). In contrast, the magnitude of increases varied substantially, averaging 294 ± 77% of the basal rate for units in SI/NB (n = 12) and 486 ± 131% in LPOA/PS (n = 13).

View larger version (12K): [in this window] [in a new window]   Fig. 5.   LPS-induced changes in firing rate as a function of neuron location. LPS-responsive neurons were defined as those demonstrating a minimum 50% change in firing rate after LPS administration compared with pre-LPS rate during same vigilance state. These criterion levels are indicated by dashed horizontal lines. Data points (shapes) denote log transform of ratios of post-LPS to pre-LPS firing rates during same vigilance state in LPS-responsive units.

Of the 74 LPS-responsive neurons, 34 (46%) responded during the 60- to 100-min postinjection interval, 18 (24%) during the 150- to 210-min interval, and 22 (30%) during both intervals (Table 4). Thus the majority (76%) of the LPS-responsive neurons showed early-onset alterations in firing rates. Twenty-three of the LPS-responsive units (31%) met criteria for LPS responsiveness only during waking, 35 (47%) responded only during sleep, and 16 (22%) responded during both vigilance states (Table 4).

Basal firing rate of LPS-responsive units in LPOA/PS was significantly slower during quiet wakefulness than were those of nonresponsive units (3.9 ± 1.3 vs. 9.2 ± 2.8, respectively; P = 0.019, Mann-Whitney rank sum test; Table 5). However, in other structures studied, basal firing rates were similar in responsive and nonresponsive units (Table 5). LPS administration was not associated with obvious changes in firing patterns; autocorrelograms were similar before and after LPS administration (data not shown).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

These experiments characterized the effects of peripheral administration of bacterial somnogen LPS on neuronal activity in brain regions classically associated with sleep and wakefulness. LPS administration was associated with altered firing rates in 39% of the neurons examined. Many of these units also showed vigilance-related alterations in firing rates. Characteristics of LPS responsiveness and vigilance-related alterations in firing rates were significantly related. LPS-related reductions in firing rates were similar in magnitude across brain regions, averaging 32% and 33% of the pre-LPS values in SI/NB and LPOA/PS, respectively. LPS-related increases in firing rates were more variable across regions, averaging 294% in SI/NB and 486% in LPOA/PS.

The comparison of pre- and post-treatment neuronal firing rates during matched behavioral states is an experimental approach known as behavioral clamping (13, 14). We applied this strategy by comparing neuronal firing rates during qualitatively similar states of sleep and quiet wakefulness that occurred before and after administration of LPS. The validity of this approach relies on the absence of qualitative differences in the behaviors performed by the animal during pre- and postdrug recording intervals. We used two strategies to minimize the likelihood that qualitative changes in physiology or behavior could account for observed changes in firing rate. First, we recorded only during discrete epochs in which the animal maintained a consistent behavioral condition (non-REMS or quiet wakefulness). States of active wakefulness (i.e., intervals during which the rats engaged in overt activities such as locomotion, rearing, grooming, or ingestion) were excluded because accurate matching of active states can be difficult to achieve; animals that are sleeping are clearly not engaging in overt movement; and neuronal activity measured during behaviorally active states may be related to specific movements, the cognitive state, or other behavioral or physiological variables, rather than to the specific condition of simply being awake. The proportions of sleep-active and wake-active neurons detected in our study (7% and 12%, respectively, across all brain regions examined) are somewhat lower than proportions previously reported in corresponding brain regions of unanesthetized rats (~24-28% sleep-active and 14-39% wake-active neurons) (3, 17, 36). This difference in proportions could be related to differences in the state-defining behavioral criteria (e.g., our exclusion of active wakefulness), minimal criteria applied to declare a neuron as vigilance-related (e.g., we required a minimal 50% change in firing rate), the type of electrode used, and other similar experimental variables.

Our second strategy for achieving accurate behavioral clamping was to conduct recording sessions during a postinjection time interval that preceded the onset of overt LPS-induced changes in sleep or temperature. In our study, administration of LPS early in the light phase of the circadian cycle induced alterations in sleep and temperature after a latency of 4-5 h. The delayed and modest physiological responses that we observed after LPS administration were not surprising, as others have previously reported that the somnogenic and pyrogenic impact of LPS and interleukin-1 is under circadian control (18, 27, 28). Typically, administration of these somnogens during the light phase of the circadian cycle, when spontaneous sleep time is normally high and temperature is normally low, elicits a robust fever and increased DWA during SWS but little or no change in total SWS time. In contrast, administration during the dark phase, when sleep is relatively low and temperature is high, elicits modest fevers and little change in DWA during SWS but marked increases in the amount of time spent in SWS.

Alterations in neuronal firing rates were detected during the initial 3.5 h after LPS administration. This time interval corresponds to intervals reported for initial and maximal c-fos expression after LPS treatment (15, 19) but precedes the onset of fever and of overt changes in SWS time and DWA during SWS (Fig. 5). c-fos gene expression is considered to reflect neuronal activation in a number of model systems (23, 32, 33). Intraperitoneal injection of LPS elicits low-to-moderate increases in c-fos expression in brain at 1 h postinjection, with maximal increases present at 3 h postinjection (19). In our study, the majority (76%) of LPS-responsive units demonstrated changes in firing rate during the 60- to 100-min period after injection. Thus the temporal pattern of LPS-induced c-fos expression in brain is consistent with our time course for detectable alterations in firing rates. Similarly, the large increases in neuronal firing rates that we observed in the LPOA during sleep are consistent with increased neuronal c-Fos immunoreactivity in the ventral LPOA during sleep (31) and corroborate the recent electrophysiological findings of others demonstrating vigilance-related alterations in neuronal firing rates in the ventral LPOA (36). However, the relationship among the variables of sleep, c-fos expression, and altered neuronal firing rates in the LPOA after LPS treatment is unclear. Despite the temporal correlations between LPS administration and c-fos expression in brain and between sleep and c-fos expression in the LPOA, increased c-fos expression in the LPOA is not a typically reported correlate of LPS treatment in rats (15, 19). Determining whether LPS-associated changes in neuronal firing rates and c-fos expression are causally related to LPS-induced alterations in sleep will clearly require further study.

Lesion and transection studies consistently demonstrate that damage to the BF or its connections is associated with reduced spontaneous sleep in experimental animals (21, 37), supporting the hypothesis that BF neuronal circuitry is involved in the regulation of sleep and wakefulness. State-dependent alterations in BF neuronal firing rates that are similar in proportion and magnitude to those reported here have been described previously in cats, rats, and dogs (3, 17, 25, 35, 37, 38). Like LPS, other pharmacological manipulations that alter sleep state development also induce changes in state-related neuronal firing rates in BF (25). However, BF, like the LPOA, is not typically noted for marked changes in c-fos expression after LPS administration (8, 145 19, 40). Instead, LPS administration elicits robust c-fos expression in nucleus tractus solitarius, the paraventricular nucleus, the bed nucleus of the stria terminalis, and the central nucleus of the amygdala (8, 15, 19, 40). Increased c-fos expression in these structures probably reflects LPS-induced activation of the hypothalamic-pituitary-adrenal (HPA) axis and the autonomic nervous system (30). In contrast to the robust LPS-induced expression of c-fos in brain structures with presumed autonomic and HPA functions, cells participating in sleep regulation may be diffusely distributed throughout structurally and functionally heterogeneous brain regions like BF (36). Indeed, sporadic c-fos-positive cells have been detected in rat BF after the administration of the somnogenic cytokine interleukin-1 (9).

Subpopulations of up to 45% of the neurons in various hypothalamic nuclei and in the horizontal diagonal band alter their firing rates in response to 1.0-1.5°C changes in temperature (1, 3, 36). Thus alterations in body temperature, such as fever, could potentially influence the firing rates of so-called warm-sensitive neurons in these areas. Warm-sensitive neurons often demonstrate increased firing rates during SWS (1), and preoptic/anterior hypothalamic warming can alter neuronal activity in the BF (2). Changes in core temperature due to LPS-induced fever could therefore potentially contribute to alterations in the firing rates of BF and hypothalamic neurons after LPS administration. In our paradigm, however, altered neuronal firing rates temporally preceded the expected onset of LPS-induced fever, as does LPS-induced c-fos expression. Moreover, most neurons that responded to LPS also showed vigilance-related changes in firing rates, suggesting that their LPS-related responses were associated with vigilance state rather than with temperature or autonomic variables. Thus the LPS-induced alterations in neuronal firing rate that we observed may reflect the initiation of physiological changes that eventually cause alterations in somnolence.

Nonetheless, a role for these neurons in the initiation of fever and autonomic responses remains a possibility. Thermoregulatory changes elicited by LPS could develop during the so-called "chill phase" that precedes the actual onset of fever. This state is associated with the activation of heat production mechanisms and with physiological and behavioral adjustments to reduce heat loss. Many cold-sensitive neurons in the rostral hypothalamus and some basal forebrain regions are wake related (1, 2, 3), and in our study, the majority (18 of 27) of vigilance-related LPS-responsive neurons were wake related (Table 3). Thermosensitive neurons are typically identified based on their response to a thermo-induced change in regional brain temperature (1, 2, 3), but cold-sensitive neurons could also potentially be activated by the difference between the actual core temperature and the elevated physiological temperature set point that triggers the generation of fever. Thus LPS-induced activation of wake-related cells could reflect activation of cold-sensitive neurons that trigger the heat production mechanisms (e.g., shivering) or heat-conservation mechanisms (e.g., piloerection) that precede and generate fever.

In conclusion, our findings are consistent with previous studies demonstrating vigilance-related changes in neuronal activity in various regions of BF and hypothalamus and, furthermore, show that LPS treatment significantly alters neuronal firing rates in these structures during both sleep and wakefulness. These changes in firing rate precede overt changes in physiological or behavioral parameters, but are temporally consistent with reported changes in LPS-related c-fos expression in brain. They could then represent a mechanism for the initiation of LPS-induced physiological changes. At present, little is known about the effects of peripherally administered somnogens at the neuronal level, and data are not available to demonstrate which endogenous somnogens impact behaviorally relevant neuronal systems under relevant pathophysiological conditions. Elucidating the mechanisms by which peripheral somnogens modulate sleep under normal or pathological conditions requires further evaluation of the neurophysiological effects of somnogenic treatments at brain sites critically associated with sleep induction and arousal.