The suprachiasmatic nucleus (SCN) regulates the circadian rhythms of body temperature (Tb) and vigilance states in mammals. We studied rats in which circadian rhythmicity was abolished after SCN lesions (SCNx rats) to investigate the association between the ultradian rhythms of sleep-wake states and brain temperature (Tbr), which are exposed after lesions. Ultradian rhythms of Tbr (mean period: 3.6 h) and sleep were closely associated in SCNx rats. Within each ultradian cycle, nonrapid eye movement (NREM) sleep was initiated 5 ± 1 min after Tbr peaks, after which temperature continued a slow decline (0.02 ± 0.006°C/min) until it reached a minimum. Sleep and slow wave activity (SWA), an index of sleep intensity, were associated with declining temperature. Cross-correlation analysis revealed that the rhythm of Tbr preceded that of SWA by 2–10 min. We also investigated the thermoregulatory and sleep-wake responses of SCNx rats and controls to mild ambient cooling (18°C) and warming (30°C) over 24-h periods. SCNx rats and controls responded similarly to changes in ambient temperature. Cooling decreased REM sleep and increased wake. Warming increased Tbr, blunted the amplitude of ultradian Tbr rhythms, and increased the number of transitions into NREM sleep. SCNx rats and controls had similar percentages of NREM sleep, REM sleep, and wake, as well as the same average Tb within each 24-h period. Our results suggest that, in rats, the SCN modulates the timing but not the amount of sleep or the homeostatic control of sleep-wake states or Tb during deviations in ambient temperature.
- brain temperature
- ambient temperature
- rapid eye movement sleep
- slow wave activity
the suprachiasmatic nucleus (SCN) of the hypothalamus drives the circadian rhythm of body temperature (Tb) and sleep-wake activity in mammals (8, 17, 28). Most studies have shown that lesioning the SCN abolishes the circadian rhythm of Tb in rats and hamsters (8, 33, 37) and the sleep-wake rhythm in rats (8, 17, 27, 45), mice (9, 18), and squirrel monkeys (10).
In the circadian rhythmic animal, sleep and brain and body temperature regulation have a close relationship. Brain temperature (Tbr) drops in association with non-rapid eye movement (NREM) sleep and rises in association with rapid eye movement (REM) sleep and wake (reviewed in Ref. 15). In humans, the peak in sleep propensity coincides with the nocturnal minimum of Tb (6), and a decline in Tb, associated with peripheral heat loss, promotes sleep onset (19). This synchrony between body temperature and sleep regulation may partly be a consequence of them both being controlled by a common pacemaker, the SCN.
Sleep and Tb also might have a homeostatic relationship that exists independently of circadian rhythmicity. Rats with lesions of the SCN (SCNx rats) display a robust ultradian rhythm in both Tb and sleep-wake states with a period of <6 h (8). Sleep and temperature appear to maintain their association within the ultradian rhythm of the SCNx rat. Eastman et al. (8) reported that ultradian variations in temperature and wake are parallel in SCNx rats, and Edgar et al. (10) reported that SCNx squirrel monkeys show an ultradian rhythm in sleep-wake states that corresponds with that of Tbr. The inverse relationship between ultradian rhythms of slow wave activity (SWA) and Tb also is apparent in rats that have reduced circadian rhythmicity following lesions of the ventral subparaventricular zone, a projection site of the SCN (22). However, no study has ever investigated the extent of the association between Tbr and sleep, independent of circadian rhythmicity. The primary objective of our study, therefore, was to thoroughly investigate the relationship between Tbr and sleep within the ultradian rhythms of the SCNx rat.
Recent studies have suggested that the SCN not only regulates the circadian distribution of sleep but also may be involved in regulating the amount of sleep in squirrel monkeys (10), mice (9), and rats (25). Most studies, however, have found no change in daily sleep time after SCN lesions in rats (16, 40, 45). Also, SCNx rats (27, 39) and hamsters rendered arrhythmic by a phase shift of the light-dark cycle (20) show a normal recovery sleep response after sleep deprivation. Mice with disrupted mPeriod genes have an amount of sleep similar to that of wild-type mice (35). Likewise, SCNx hamsters maintain a similar mean 24-h Tb, although daily variation in Tb is decreased, compared with intact animals (32). Also, SCN-lesioned rats (43) and hamsters (32) exposed to extreme ambient temperature (Ta) still exhibit normal thermal homeostasis. These studies suggest that sleep and temperature homeostasis remains intact in rodents even if circadian rhythmicity is eliminated.
Sleep-wake states are sensitive to a change in Ta: rats exposed to extreme cold or hot sleep less; however, mild warming increases NREM and REM sleep (15, 24). Whether the rats' responses to Ta are influenced by the circadian system is unknown. The second objective of this study, therefore, was to determine whether SCNx rats are able to thermoregulate and regulate their sleep as efficiently as intact rats when challenged by 24-h exposure to Ta that deviated from thermoneutrality. By evaluating sleep-wake states and Tb over 24 h, we also were able to investigate further the role of the SCN in regulating daily sleep amount and mean Tb in rats.
We report that Tbr and sleep-wake states have a close relationship independent of circadian rhythmicity in rats maintained at room temperature. We also report that in the rat, the SCN regulates the distribution of sleep-wake states, but not sleep amount, regardless of Ta and in the absence of light cues.
MATERIALS AND METHODS
Twenty-four male Sprague-Dawley rats, weighing 280–350 g at the beginning of the experiment, were used. All experimental procedures were approved by the Animal Care and Use Committee at Veterans Affairs Greater Los Angeles Healthcare System and were conducted according to the guidelines of the National Research Council. Rats underwent only one surgical procedure. They were anesthetized with ketamine/xylazine (80/10 mg/kg ip). A radio transmitter for Tb and locomotor activity monitoring (PDT 4000 E-mitter; Mini Mitter, Bend, OR) was surgically placed in the peritoneal cavity of each rat. Next, SCN lesions were made in 17 rats with a radio frequency lesion maker (model LM4; Grass Medical Instruments, Quincy, MA). After rats were secured in a stereotaxic frame (Kopf Instruments, Tujunga, CA), a Tungsten electrode (diameter 0.25 mm; A-M Systems, Carlsborg, WA), insulated except for the tip, was lowered four times through a small craniotomy targeting the SCN at the following coordinates: anterior-posterior (AP) −0.6 and −1.1 mm from bregma, lateral (L) ±0.2 mm from superior saggital sinus, and dorsoventral (DV) 8.9 mm from the dura (31). Lesions were made by increasing the current to 2–3 mA for 60 s at each lesion site. The electrode was lowered, but no current was applied to make sham lesions in seven control rats. Finally, all rats were surgically implanted with chronic cortical EEG and dorsal neck electromyogram (EMG) electrodes for determining arousal state. Briefly, four stainless steel screw electrodes were threaded into the skull (AP +2 mm, L ±2 mm, AP −4 mm, and L ±3 mm from bregma) for EEG recordings, and flexible, insulated stainless steel wires were inserted into dorsal neck muscles for EMG recordings. A stainless steel guide tube (outer diameter 0.6 mm) was stereotaxically placed just above the dorsal margin of the hypothalamic lateral preoptic area/basal forebrain at coordinates AP −1 mm from bregma, L −2 mm from bregma, and DV −5.5 mm below the skull, through which a thermistor (bead diameter 0.41 mm; Thermometrics, Edison, NJ) was inserted for recording Tbr. Leads from the electrodes and the thermistor were soldered to a small Amphenol connector, and the complete assembly was anchored to the skull with dental acrylic.
Rats were allowed a 1- to 2-wk recovery period after surgery, during which time they were housed in individual cages, in a temperature-controlled recording chamber (23 ± 0.6°C), with lights on at 0700. Rats had ad libitum access to food and water. After recovery, all rats were kept in constant dim red light (4–6 lux) for a 2-wk screening period, during which time Tb and locomotor activity were continuously recorded through a radio receiver (ER-4000; Mini Mitter) placed underneath each cage. Data were recorded and stored in 2-min bins using the VitalView data acquisition system (Mini Mitter). Rats were maintained in constant dim red light conditions throughout the screening and recording period because light-dark cycles can influence sleep patterns independently of the circadian timing system (26).
For the last 3–4 days of the screening period, rats were adapted to recording cables. After adaptation, rats were exposed to three Ta (24, 18, and 30°C), each for a 24-h period separated by a 24-h recovery period. Order of exposure to the different Ta was randomized between rats (recorded in pairs). Ta was continuously monitored using data loggers (HOBO H8; Onset Computer, Pocasset, MA). Average recorded chamber temperatures were 23.1 ± 0.6, 18.2 ± 0.6, and 30.2 ± 0.6°C for the three recording periods. Ta was set to the required level 90 min before the start of each recording to allow chamber temperature to equilibrate. At this time, rats also were connected to the recording cable, and they remained connected for the duration of the experiment. Recordings began at 1000.
Rats were connected to a recording cable that was lightly suspended above them by a counterweighted beam. The cable joined the miniature connector on the animal's head to a polysomnographic recording system (Embla; Medcare Flaga Medical Devices, Reykjavik, Iceland). The thermistor was connected directly to a bridge-amplifier (NTC sensor bridges; INTEC Associates, Surrey, UK), which fed into the Embla isolation unit and recorder. EEG, EMG, and Tbr recordings were digitally displayed and stored on a computer using Somnologica software (Somnologica Studio; Medcare Flaga Medical Devices). EEG and EMG signals were digitized at a sampling rate of 100 Hz. The EEG and EMG signals were filtered (EEG: low pass, 30 Hz and high pass, 0.3 Hz; EMG: low pass, 100 Hz and high pass, 10 Hz) before analysis. Thermistors were calibrated, by water immersion, against a precision thermometer (YSI 4600; Yellow Springs Instruments, Dayton, OH), to an accuracy of 0.01°C. Because of bridge circuit problems, baseline Tbr was offset in some rats so that we report change in Tbr rather than absolute temperatures.
At the end of the experiment, all rats were perfused transcardially with 0.1 M phosphate-buffered saline (PBS), followed by 300 ml of fixative containing 3% paraformaldehyde and 15% picric acid in 0.1 M PBS, followed by 10% and then 30% sucrose. The brains were then stored in 30% sucrose at 4°C until they sank. Coronal sections were cut at 40 μm on a freezing microtome. Sections were processed for vasoactive intestinal polypeptide (VIP) immunostaining to confirm that the SCN had been completely lesioned in SCNx rats; SCN neurons contain higher levels of VIP than the surrounding hypothalamic areas (5). Briefly, sections were incubated with a rabbit anti-VIP primary antiserum (1:500; Immunostar, Hudson, WI) for 48 h at −10°C. Sections were subsequently incubated with a biotinylated goat anti-rabbit IgG (1:200; Vector Laboratories) for 2 h and then reacted with avidin-biotin complex (ABC, 1:100; Vector Elite kit) and developed with diaminobenzidine tetrahydrochloride for visualization.
Ten-day periods of temperature and activity data in candidate SCNx rats were analyzed for the loss of circadian rhythms by applying the χ2-periodogram procedure (36) to collapsed 6-min bins of temperature and activity data (Circadian Physiology Software (http://www.circadian.org/biorhyt.html), version 4.0). We looked for periodicity in the interval from 180 to 280 bins (i.e., from 18.0 to 28.0 h, with 0.1-h resolution) and used the value of the Sokolove-Bushell's Qp (36) as an index of robustness of the rhythms. Only rats with verified circadian arrhythmicity and absent VIP staining of the SCN (n = 7) were included in the SCN-lesioned group (SCNx) for final analysis. Ten lesioned rats still showed significant circadian periodicity and incomplete lesioning of the SCN. Data from these rats were excluded from analysis. Period of the circadian temperature and activity rhythms of control rats was determined using the χ2 periodogram procedure.
Analysis of sleep-wake states.
Sleep-wake states of the rats were determined by an experienced scorer on the basis of the predominant state within each 10-s epoch. The scorer was blind to experimental condition and group identity of the animal. Wake was defined as low-voltage, high-frequency activity combined with elevated neck muscle tone. NREM sleep was defined as high-amplitude EEG with prominent activity in the 2- to 4-Hz range. REM sleep was defined as moderate-amplitude EEG with dominant theta frequency activity (6–8 Hz) combined with minimal neck EMG tonus except for occasional brief twitches.
Spectral analysis was performed on the digitized EEG signals across each 24-h recording period. The EEG recordings were subjected to a fast Fourier transform routine (Somnologica Studio), after which a power spectrum was computed for the frequency range of 0.75–4.0 Hz (SWA) for each 10-s epoch. Epochs containing artifacts were omitted from spectral analysis.
One control rat was excluded from all analyses because of incomplete sleep-wake and temperature data. The final sample therefore consisted of seven SCNx rats and six controls. Sleep-wake data were incomplete for one SCNx rat for the 18°C condition; however, this rat was kept in the sample for other analyses. Tbr data sets were complete for six SCNx and five control rats for the 24 and 18°C conditions and for five SCNx rats and four control rats for the 30°C condition. Tb data were missing for one control and one SCNx rat.
Statistical procedures were performed using Statistica (StatSoft, Tulsa, OK). All results are reported as means ± SE.
Association between Tbr and vigilance states.
The 24°C recording was used to explore the characteristics of the Tbr and sleep patterns, and their associations, in the SCNx and control rats. First, 10-s Tbr values across 24 h were smoothed with a 1-min moving average, and then χ2-periodogram analysis was performed. Significant ultradian rhythmicity of Tbr was assessed in the interval from 20 to 60 bins (i.e., from 2.0 to 6.0 h). The amplitude of each ultradian cycle was calculated as the maximum minus the minimum temperature, and values were averaged for each rat.
The temporal association between Tbr and SWA was investigated for the 24°C condition by using cross-correlation analysis (time series analysis, Statistica). Cross-correlation analysis measures the correspondence and phase relationship between two time series across a temporal range of overlap (44). For cross-correlation to be valid, both time series must be stationary and must be sampled at equal time points. SWA and Tbr time series were prepared for analysis as follows. Missing values in the SWA data, caused by artifact, were replaced by using linear interpolation, and SWA data were normalized by taking the natural log of each value. Ten-second epochs of Tbr and the natural log of SWA then were pooled into 1-min epochs before further smoothing with a 5-min moving average. Data were detrended by subtracting the mean of the time series from each data point, and any remaining linear trend was removed using linear regression. Two SCNx rats and one control rat were excluded from analysis because of missing temperature data (n = 2) or excessive artifact in SWA data (n = 1). Correlations were therefore investigated for five SCNx and five control rats. Correlations between Tbr and SWA at lag times of between 0 and 60 min were explored. SWA was used as the lagged variable. For positive lags, the Tbr rhythm preceded the SWA rhythm. Significance of correlation coefficients was determined with a t-test. A t-test then was used to compare average peak correlation coefficients (that had been transformed with the Fisher z transform) between the two groups of rats.
Detailed analysis also was made of the change in SWA in association with declining Tbr for each complete ultradian temperature cycle in each rat, using the same smoothed data described above. SWA was averaged across the descending limb of each ultradian Tbr cycle as follows. Temperature peaks and troughs were visually identified, and the number of 10-s epochs between these two points was calculated and divided into deciles. SWA and Tbr values were then averaged within each decile across the descending temperature curve for each rat and then for each group; values are reported as a change from the minimum temperature and the corresponding SWA in that interval. An average of six ultradian temperature cycles were identified and used in the analysis for each rat. Although the temperature peaks and troughs were less regular in control rats compared with SCNx rats and were superimposed on the circadian rhythm, they were identifiable.
Finally, we calculated the rate of change of Tbr from the maximum Tbr to sleep onset (first continuous minute of NREM sleep) and from sleep onset to the minimum Tbr for each ultradian temperature cycle in each rat. Average rates of change for all ultradian temperature cycles were computed for each rat and subsequently averaged for each group. In addition, the average time periods between maximum temperature and sleep onset and between sleep onset and minimum temperature for each ultradian temperature cycle were calculated and averaged for each group of rats.
Effect of Ta on Tbr, Tb, and vigilance states.
The effect of three different Ta settings on Tb, Tbr, and vigilance states in SCNx and control rats was investigated in other analyses. Control rats had free-running rhythms in the dim light conditions, and we did not align their rhythms to circadian time. Rather, all results are reported on the basis of the duration of exposure to each Ta, starting at 1000. Differences in 24-h averages for each vigilance state and Tb were investigated according to Ta and experimental group (SCNx vs. control) by using a two-way repeated-measures ANOVA, followed by Fisher's exact test when appropriate. For Tbr analysis, t-tests for a single mean were used to investigate whether the 24-h average difference in temperature from that at 24°C was significant for either the 18 or 30°C condition for each group of rats. Differences in Tbr for the 18 and 30°C conditions also were analyzed separately according to vigilance state by using t-tests for a single mean.
For analysis of the distribution of vigilance states, SWA, and Tb across 24 h during each experimental condition, 2-h values were computed and subjected to a two-way, repeated-measures ANOVA, with group (SCNx or control) as the “between factor” and Ta (24, 30, or 18°C) and time (2-h intervals) as “within factors.” ANOVA was followed by a Fisher exact post hoc test to find the origin of significance where appropriate. For standardization, values for SWA were expressed as percentages of the average SWA in NREM sleep over the 24-h recording for the 24°C condition. Data were then transformed using the natural log before statistical analysis was performed.
To characterize the relationship between vigilance states and Tbr at different Ta, the change in Tbr during transitions between vigilance states was investigated according to the method previously described in detail by Gao et al. (12). The following transitions were investigated: 1) wake-NREM, 2) NREM-REM, and 3) NREM-wake. Briefly, 4-min episodes containing a clear transition from one vigilance state (VS1) to another (VS2) were selected. In the 2 min preceding or the 2 min following the transition, 80% or more had to be scored as VS1 or VS2, respectively. Furthermore, the three epochs preceding the transition had to be VS1, and the first epoch after the transition had to be VS2. In the selected 4-min episodes, 10-s values of Tbr were determined, excluding any epochs not corresponding to VS1 or VS2 before and after the transition, respectively. Within each 4-min episode, 10-s values were expressed relative to the value at the transition (calculated as the mean of the value in the last VS1 epoch and the first VS2 epoch). Temperature was expressed as the difference from the transition value. Mean transitions were first calculated for each individual for each 24-h recording period, and then mean transition curves were constructed for each group of rats for each 24-h experimental condition. For statistical analysis, we used a two-way, repeated-measures ANOVA followed by the Fisher post hoc test when appropriate to compare the change in Tbr after each transition (data averaged for the first and second minute) between groups and according to Ta. Finally, the numbers of each transition within each 24-h period were calculated and statistically compared using ANOVA as described above.
Figure 1 shows the χ2-periodogram analyses of Tb from a representative SCNx rat and a control rat. The SCN was successfully lesioned in the seven SCNx rats, as indicated by the absence of significant circadian periodicity in Tb and activity during 10 days in dim light. All control rats had free-running circadian periodicities >24 h for Tb (range: 24.0–24.9 h) and activity rhythms (range: 24.1–25.2 h). Photomicrographs of sections through the SCN in representative SCNx and control rats are shown in Fig. 2. VIP immunostaining of SCN cells, clearly visible in the control rats, was absent in all SCNx rats, confirming the successful placement of the lesions. All lesions damaged the adjacent periventricular nucleus and ventromedial part of the medial preoptic nucleus. In some but not all rats, the lesions damaged the optic chiasm, anteroventral periventricular nucleus, retrochiasmatic area, ventromedial anterior hypothalamic area, and, in single cases, limited parts of the arcuate or ventromedial paraventricular nuclei.
Relationship Between Tbr and Vigilance States
Tbr, Tb, SWA, and hypnograms from representative SCNx and control rats during their 24-h recording (Ta = 24°C) are shown in Fig. 3. SCNx rats showed no circadian rhythmicity in either temperature or sleep-wake states and, instead, displayed a regular ultradian rhythmicity, most evident in their Tb and Tbr. χ2-periodogram analysis of Tbr revealed that SCNx rats had a significant ultradian rhythmicity of 3.6 ± 0.8 h. All control rats had a circadian rhythm, and in addition, four control rats showed significant ultradian periodicity between 2.8 and 5.6 h. The remaining two control rats did not have a significant ultradian rhythm. The average amplitude (maximum minus minimum temperature) of the ultradian temperature cycles in the SCNx and control rats was similar (SCNx: 1.2 ± 0.08°C; controls: 1.1 ± 0.06°C).
Also evident from Fig. 3, and shown in an expanded view in another SCNx rat in Fig. 4A, is the coincidence of high SWA and temperature troughs and low SWA and temperature peaks. This inverse relationship was statistically confirmed with cross-correlation analysis. The peak correlation coefficients between SWA and Tbr and significant time lags are shown for each rat in Table 1. Figure 4B shows a plot of correlation coefficient vs. time lag for one SCNx rat (X1), illustrating the association between Tbr and SWA (lagged variable). Four SCNx rats and four control rats showed a significant negative correlation between Tbr and SWA at a positive time lag of ≤10 min. In other words, the ultradian rhythms of Tbr and SWA would be maximally negatively correlated if the rhythm of SWA were shifted forward by 2–10 min. Average z-transformed correlation coefficients were not significantly different (P = 0.9) between SCNx (−0.39) and control rats (−0.46).
A closer inspection of the change in SWA in association with the declining temperature limb of the ultradian temperature cycles for SCNx and control rats further illustrates the inverse relationship between SWA and Tbr (Fig. 5). SWA, a marker of NREM sleep intensity (3), rises sharply as Tbr begins to decline and remains relatively high until minimum Tbr is reached. SWA varied between ultradian curves and between rats, depending on the amount of intervening REM sleep and wake; however, the declining portion of the temperature curve was associated mostly with NREM sleep, whereas REM sleep mostly occurred near the temperature minima and wake occurred mostly in association with the rising portion of the curve (data not shown).
Analysis of the rate of change in Tbr for each ultradian temperature cycle revealed that Tbr dropped rapidly (SCNx rats: 0.07 ± 0.01°C/min; controls: 0.06 ± 0.004°C/min) over a short period (SCNx rats: 5 ± 1 min; controls: 6 ± 1 min) before the onset of the first minute of NREM sleep. In support of the cross-correlation analysis, therefore, sleep onset lags behind the drop in Tbr by a few minutes. Tbr continued to decline more gradually after sleep onset (SCNx rats: 0.02 ± 0.006°C/min; controls: 0.02 ± 0.004°C/min) over a longer period (SCNx rats: 61 ± 22 min; controls: 56 ± 7 min) that consisted mostly of NREM sleep until the minimum temperature was reached.
Influence of Ta on Tb, Tbr, and Sleep-Wake States
Effects of Ta on Tb and vigilance states for each 24-h period of exposure to 24, 18, and 30°C are shown in Tables 2 and 3. Ta had no significant effect on 24-h Tb or on minimum or maximum Tb. Tbr was more sensitive to Ta: SCNx rats had a significantly increased average 24-h Tbr at both 18 and 30°C compared with 24°C (Table 2). Control rats tended to have a higher 24-h Tbr at 30°C compared with 24°C (P = 0.09), although not significantly. When change in Tbr was analyzed according to vigilance state, significant effects of Ta were apparent. At 18°C, SCNx rats had significantly higher Tbr during wake (P = 0.01) and tended to have higher Tbr in NREM and REM sleep (P = 0.07). The change in Tbr was not significant in the control rats according to vigilance state for the 18°C condition. At 30°C, both groups of rats had Tbr that were significantly higher than those at 24°C during NREM sleep (P < 0.05) but not during wake (P = 0.1). SCNx rats also had higher Tbr during REM sleep at 30°C (P = 0.02), which appeared only as a tendency for controls (P = 0.09), probably because of the small sample size. Finally, the average amplitude of the ultradian temperature cycles was significantly blunted in the 30°C condition regardless of group [Ta effect: F(2,14) = 30.4, P < 0.0001; Fisher's post hoc P < 0.0001].
Ambient cooling, but not ambient warming, affected average percentage of time spent in vigilance states over 24 h. SCNx and control rats spent significantly more time awake and less time in REM sleep during 24 h at 18°C compared with both 24 and 30°C (Table 3).
Figures 6 and 7 show the average Tb and Tbr and sleep-wake state distribution, at 2-h intervals, across 24 h of exposure to the three Ta values for the SCNx and control rats. There were significant group × time interactions for Tb [F(11,99) = 5.6, P < 0.0001], NREM sleep [F(11,110) = 4.2, P < 0.0001], REM sleep [F(11,110) = 3.8, P < 0.001], SWA [F(11,99) = 2.7, P = 0.004], and wake [F(11,110) = 4.3, P < 0.0001]. Post hoc analysis revealed no significant differences among any time points across 24 h for any of the variables for SCNx rats. In contrast, control rats displayed a circadian rhythm in all variables regardless of Ta. They had significantly lower Tb values and spent significantly more time in NREM and REM sleep between 1400 and 2200 than between 0200 and 1000 (Fisher's post hoc, P < 0.04). SWA was significantly higher during the first 4 h of the major sleep period (1400–1800) compared with the time period from 2000 to 0600 (Fisher's post hoc, P < 0.04). Finally, percentage wake was significantly lower between 1400 and 2200 than between 0200 and 0800 (Fisher's post hoc, P < 0.02). There also was a significant difference in percentage wake between SCNx rats and controls at 0600–0800, with controls spending more time awake compared with SCNx rats (Fisher's post hoc, P = 0.03). Controls tended to spend less time awake at 1600–1800 compared with SCNx rats (P = 0.1).
Regardless of group, Ta had a significant effect on Tb that interacted with the factor time [F(22,198) = 2.8, P < 0.001]. Tb was higher during the first 2 h and during the last 8 h of the 30°C compared with the 24°C recording (Fig. 6). Tb was sporadically higher during the 18°C compared with the 24°C recording (Fig. 6). There were no significant group or group × Ta interaction effects for Tb (P > 0.9).
There was a significant Ta effect for NREM sleep [F(2,20) = 5.0, P = 0.02] but no significant post hoc comparisons and no significant Ta × time interaction (P > 0.5). REM sleep, however, was significantly affected by Ta, which interacted with the factor time [F(22,220) = 2.6, P < 0.001]. Percentage REM sleep was lower during the first 2 h and again 8–10 h later for the 18°C compared with the 24°C recording (Fig. 7). In contrast, REM sleep was significantly higher 6–8 h into the 30°C compared with the 24°C recording (Fig. 7). Repeated-measures ANOVA revealed a significant Ta effect on percentage time spent awake [F(2,20) = 11.6, P = 0.0005], but there were no significant post hoc comparisons, probably because of some Ta × time interactions, although the overall interaction effect was not significant [F(22,220) = 1.3, P = 0.2]. Finally, there also was a significant Ta effect [F(2,18) = 5.6, P = 0.01] and Ta × group effect [F(2,18) = 6.1, P = 0.009] for SWA, but there were no significant post hoc comparisons, possibly because of some spurious Ta × time × group interactions, even though this interaction effect was not significant [F(22,198) = 1.3, P = 0.15].
Figure 8 shows the average 24-h change in Tbr at transitions between vigilance states at different Ta. Temperature decreased when a period of NREM sleep was entered and increased when REM sleep was entered or during wake in both groups of animals. The temperature decline after NREM sleep was entered from wake was blunted in the 30°C condition, regardless of group [Ta effect: F(2,14) = 8.7, P = 0.003]. On waking from NREM sleep at 30°C, the Tbr increase also was blunted in the second minute [F(2,14) = 13.4, P < 0.001]. After the transition from NREM to REM sleep, the increase in Tbr was significantly less at 30°C but greater at 18°C compared with that at 24°C [F(2,14) = 28.5, P < 0.001] (Fig. 8). Table 4 shows the number of transitions between wake and NREM sleep, NREM and REM sleep, and NREM and wake. Repeated-measures ANOVA revealed a significant Ta effect on the number of wake-NREM sleep transitions [F(2,14) = 6.3, P = 0.01], which was due to a significantly higher number of transitions in the 30°C compared with the 24°C condition,regardless of group (Table 4). There also was a significant Ta effect on the number of transitions from NREM to REM sleep [F(2,14) = 7.9, P = 0.005], which was reduced in the 18°C condition. Finally, there was an almost significant Ta effect for the number of NREM-wake transitions [F(2,14) = 3.6, P = 0.055], which was higher in the 30°C compared with the 24°C condition (Fisher's post hoc, P = 0.02).
Our analyses show the existence of a close relationship between sleep parameters and Tbr in the absence of circadian rhythms in SCN-lesioned rats. In SCNx rats, SWA was negatively correlated with Tbr. Within the ultradian temperature cycles that are exposed in SCNx rats, a rapid drop in Tbr heralded the onset of sleep a few minutes later, after which Tbr continued a slow decline in association with NREM sleep. As Tbr fell, there was a corresponding rise in SWA, which peaked at minimum Tbr. This relationship between SWA and Tbr was most obvious in SCNx rats but also was found in the ultradian cycles of sham-lesioned control rats.
Our results also confirm the remarkable stability of mean Tbr and Tb, as well as sleep-wake states across 24 h in rats in which circadian rhythmicity is absent. SCNx rats had the same 24-h mean Tb, percentage NREM sleep, SWA within NREM sleep, REM sleep, and wake compared with control rats. However, sleep-wake states were distributed evenly across 24 h as opposed to the more consolidated sleep and wake periods characteristic of the normal rat. Also, rats with SCN lesions were able to thermoregulate as efficiently as control rats at mildly increased or decreased Ta such that their sleep was only minimally disrupted by the change in Ta. Our results therefore support the hypothesis that, in rats, loss of circadian rhythmicity through ablation of the SCN does not disturb sleep or temperature homeostasis.
We chose to use the SCNx rat to explore the association between vigilance states and Tb to exclude circadian rhythmicity as a confounding variable. Indeed, circadian rhythms were successfully ablated in the SCNx rats in our study, and a robust ultradian rhythm in temperature became apparent with a period of 3.6 h, similar to that reported for locomotor activity in SCNx mice (3.86 ± 0.8 h) (9) and to that of Tbr in SCNx squirrel monkeys (3.8 h) (10).
Association Between Tbr and Vigilance States
As far as we are aware, our study is the first to make a detailed analysis of the ultradian rhythms of Tbr and sleep-wake states, showing their close association in the absence of circadian rhythmicity. It is well known that the circadian rhythms of sleep-wake states and temperature are closely associated. For example, rats sleep mostly during the day when Tb values are lower. In humans, under conditions of forced desynchronization, latency to sleep onset is shortest and sleep propensity is highest around the Tb minimum and sleep is terminated on the ascending portion of the Tb curve (6). Sleep initiation and sleep duration are therefore influenced by circadian phase, as measured using the temperature rhythm. Our results in SCNx rats show that sleep and Tb not only are associated because of a common circadian output but also remain tightly coupled in the absence of circadian control. A rapid decline in Tbr always precedes the onset of major sleep episodes, which are distributed regularly across 24 h. Others also have reported a decline in Tbr, beginning approximately 8 min before NREM sleep onset in rats with circadian rhythms (2). This initial decline in Tbr before EEG-defined sleep onset may be due to behavioral changes, such as reduced motor activity (12).
We found that the rhythms of Tbr and SWA were maximally correlated (negatively) at a positive time lag (2–10 min), with SWA lagging behind Tbr. The relationship between the two rhythms accounts for ∼15–20% (r2 × 100) of the variance. Our results support the hypothesis that Tb and sleep are not only kept in synchrony because of a common circadian output but that a decline in Tb may trigger sleep (14). Similar to our results, an increase in skin temperature and peripheral heat loss and a reduction in core temperature anticipates sleep onset in humans (19), regardless of whether sleep is initiated at the usual bedtime or delayed by 2 h (41).
In a separate analysis of, on average, 25 wake-NREM transitions per animal, we did not find a decrease in Tbr over the immediate 2-min period preceding NREM sleep (Fig. 8), likely because of the inclusion of transitions that were preceded by relatively short periods of wake. However, Gao et al. (12) found a decline in Tbr in the 2-min period preceding sleep, especially in the dark phase when rats were most active. Possibly, the constant dim light conditions in our study may have influenced the extent of the decline in Tbr at the wake-NREM transition.
We found similar relationships between body temperature rhythms and vigilance states in control rats. Because light can influence sleep independently of the circadian system (26), we conducted all our experiments in dim light. A consequence of this approach was that the intact rats had more prominent ultradian temperature rhythms than under conditions of light-dark and lacked a large peak in SWA, normally associated with lights on (7). Also, the control rats were free running and were at slightly different circadian phases at the start of the experiment. However, the control rats all had peak Tb values in the second half of the 24-h recording, as normally occurs during the dark phase in light-entrained rats. Thus we think that coupling of ultradian temperature rhythms and vigilance states is also characteristic of circadian rhythmic rats.
The mechanisms underlying the coupling of the thermoregulatory system and the sleep regulatory system are not completely understood. The preoptic area (POA) regulates both sleep and temperature (reviewed in Refs. 4, 24). Local POA warming triggers NREM sleep onset, increases the amount of NREM sleep (reviewed in Ref. 24), and increases SWA within NREM sleep (23). Sleep therefore may be modulated by thermosensitive neurons in the POA (1). In support of this hypothesis, warm-sensitive neurons in the POA increase their discharge during the transition from wake to NREM sleep, starting several seconds before sleep onset, and during NREM sleep (1). Heat loss is thought to be initiated by the activation of POA warm-sensitive neurons (4). Thus activation of POA warm-sensitive neurons could account for the coupling of sleep onset, heat loss, and increased SWA in association with heat loss. However, this model predicts that increased discharge rate of POA warm-sensitive neurons would occur several minutes before sleep onset, when Tb begins to decline, but no evidence at the cellular level for this prediction has been reported. Rather, current evidence suggests that there is a delay between peripheral heat loss, the decline in Tb and Tbr, and a change in thermosensitive neuronal firing rate in the POA. Possibly, Tbr might need to decrease to a certain threshold before triggering a response in the POA sleep-promoting warm-sensitive neurons. Once this threshold was reached, warm-sensitive neurons would increase their firing rate, cold-sensitive neurons would decrease their firing rate, and sleep would be initiated. Alternatively, thermal afferents might directly influence mechanisms regulating NREM sleep; an increase in skin temperature could activate warm-sensitive neurons in the POA, which are involved in sleep regulation (14, 24, 42).
Not only does a change in Tb influence sleep, but sleep also influences temperature. As previously shown (reviewed in Ref. 15), we found a gradual decrease in Tbr after sleep onset, which occurred in association with NREM sleep, and an increase in SWA. The temperature decline during NREM sleep may result from a decreased set point for heat loss responses at sleep onset (15). On the other hand, during the transition from NREM sleep to REM sleep, Tbr increased sharply, as described previously (2, 12). Vigilance states therefore have a significant impact on Tbr. In fact, Franken et al. (11) concluded that as much as 84% of the daily change in Tbr in the rat is due to sleep-wake transitions, with only 16% of the variance being accounted for by circadian influences.
Influence of SCN Lesions on Sleep and Tb Regulation
Similar to results of most studies of SCNx rats, we have found that destruction of the SCN alters the distribution of sleep across 24 h but does not influence sleep amount (17, 27, 28, 40, 45). Percentage of NREM and REM sleep as well as percentage of wake was similar in SCNx and control rats in our study. Furthermore, a thermal challenge had a similar effect on sleep-wake states in SCNx and control rats. Some recent studies, however, have found that SCNx animals not only have an altered distribution of sleep-wake states but also have increased sleep over a 24-h or circadian equivalent period. SCNx squirrel monkeys have a substantial (∼4.0 h) increase in light slow wave sleep (SWS), with no change in deep SWS or REM sleep (10). SCNx mice were found to have a large increase in NREM sleep (∼2.0 h) compared with intact mice (9), although an earlier study in mice found no difference in amount of SWS after SCNx lesions (18). Finally, a recent study reported a small (∼26 min/24 h) increase in NREM sleep and a decrease in REM sleep, with no overall difference in total sleep time in SCNx rats compared with intact rats (25). The reasons for the contradictory findings between studies of SCNx animals, particularly among rodents, are unclear. Entrainment to lighting conditions (12:12-h light-dark), applicable to some earlier studies, may influence sleep amount (25). However, studies that used constant dim light (25, 27, 45, and this study) have still produced conflicting results. Possibly, damage to neighboring structures of the SCN may influence sleep amount; damage to the POA might oppose a sleep-increasing effect of discrete SCN lesions (10). Information about the extent of the lesions in the study by Mendelson et al. (25) is not given, preventing a comparison with our rats. However, the collateral damage from the lesions in our SCNx rats is similar to that reported for SCNx mice (9), as well as SCNx rats (45). We therefore do not believe that the SCN lesions in our rats were more extensive than in other studies. Two rats that were excluded from the SCNx group in our study because they still showed significant circadian periodicity and had only partial lesions of the SCN still had normal sleep-wake percentages in 24 h (NREM sleep: 45% and 46%; REM sleep: 8% and 10%), despite having an extent of damage to neighboring structures in the POA similar to that in the SCNx rats. Furthermore, it has been reported that lesions restricted to the ventromedial POA, the locus of our collateral damage, have no effect on total sleep time (21). These findings suggest that collateral damage in the SCNx rats did not affect the amount of sleep. Thus our findings support a role of the SCN in the consolidation and timing of sleep but not in the determination of the amount of sleep in rats.
Similar to findings of previous studies in hamsters (32) and rats (43), we also found that lesions of the SCN eliminate the circadian rhythm of Tb but do not affect the homeostatic control of Tb. There was no difference in the 24-h mean Tb at 24°C in the SCNx rats compared with controls. Furthermore, when exposed to 18 or 30°C, SCNx rats were able to defend their normal Tb as effectively as control rats. These results indicate that SCNx rats have a normal, functioning thermoregulation system.
Influence of Ta on Sleep-Wake States
Exposure to mildly elevated or reduced Ta influenced Tbr and Tb in both groups of rats. Tbr was significantly higher at 30°C during NREM sleep, but not wake, in both groups of rats, which supports previous findings that regulation of Tbr is less effective during sleep than during waking (12). At transitions between wake and NREM sleep, the typical decrease in Tbr was blunted at 30°C, as previously shown (2, 12), and the amplitude of the ultradian temperature rhythm also was blunted. Apart from this blunting effect, other Tbr changes characteristic of NREM sleep, REM sleep, and wake at 24°C persisted in the cold and warm conditions.
Exposure to a cooler Ta (18°C) had less effect on Tb and Tbr than did the warm environment. However, SCNx rats had significantly higher Tbr, especially during wake, which may be due to increased shivering, vasoconstriction, or other means of increasing metabolic heat production. Indeed, rats may begin to shiver at a Ta of 20°C (13). Control rats did not show a similar increase in Tbr in the cold condition, but sample size was small.
Changing Ta also influenced sleep-wake states in the same way in both groups of rats. REM sleep was more sensitive to changes in Ta, particularly exposure to a cool Ta, than was NREM sleep, confirming previous findings (30, 38). In the cool Ta, wake was increased and REM sleep was decreased, particularly in the first 2 h of the recording. Also, the number of transitions from NREM to REM sleep was reduced at 18°C, similar to previous findings (34). REM sleep is probably decreased when Ta falls outside a thermoneutral range, because thermosensitivity is lowest and thermoregulatory responses are inhibited during this stage (15, 29).
Mild warming also influenced REM sleep to a limited extent in our study; rats had increased REM sleep near the beginning of exposure to 30°C, which is close to the Ta at which REM sleep amount peaks (29°C) (38). REM sleep begins to decrease at 31°C and is depressed at a higher Ta (33°C) (38). Unlike some previous studies in rats, we did not find a consistent increase in NREM sleep (2) or SWA (12) during mild warming in either group of rats, which may be due to different Ta conditions, as well as different circadian times at which the warming began. For example, in the study by Gao et al. (12), Ta was changed to 30°C at dark onset, when the rats typically began their active phase, whereas our control rats were near the end of their active phase when the warming began. However, we found an increase in the number of transitions from wake to NREM sleep, similar to findings of Gao et al. (12), indicating some NREM sleep-enhancing effect of exposure to 30°C.
In conclusion, we have shown that SCNx rats are able to thermoregulate and regulate their sleep amount independently of circadian input at three different Ta conditions. These results suggest that, in rats, the SCN modulates the timing but not the amount of sleep or homeostatic regulation of sleep and temperature. Furthermore, increased SWA is coupled to the declining phase of the ultradian temperature cycles that are exposed in SCNx rats. We also have shown that NREM sleep onset occurs a few minutes after the initial decline in Tbr, independent of circadian rhythmicity, suggesting that a drop in temperature may signal sleep onset.
This research was supported by National Institute of Mental Health Grants MH-47480 and MH-004708 and the U.S. Department of Veterans Affairs Medical Research Service.
We thank Feng Xu for excellent technical assistance, Amanda Turner for assistance with data analysis, and Natalia Suntsova and Irma Gvilia for comments on the manuscript.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2005 the American Physiological Society