Authors:

Róbert Bódizs,1* Sándor Kántor,2 Gábor Szabó,3 Anna Szűcs,1 Lóránd Erőss,4 and Péter Halász1

1 Epilepsy Center, National Institute of Psychiatry and Neurology, Budapest, Hungary
2 Laboratory of Neurochemistry and Experimental Medicine, National Institute of Psychiatry and Neurology, Budapest, Hungary
3 Neurotrend, Ltd., Budapest, Hungary
4 Department of Neurosurgery, MÁV Hospital, Budapest, Hungary

*Correspondence to: Róbert Bódizs, M.Sc.,
Epilepsy Center, National Institute of Psychiatry and Neurology,
Hűvösvölgyi út 116, H-1021 Budapest, Hungary.
E-mail: bodrob@net.sote.hu

Accepted for publication 13 February 2001

 

ABSTRACT

Hippocampal rhythmic slow activity (RSA) is a wellknown electrophysiological feature of exploratory behavior, spatial cognition, and rapid eye movement (REM) sleep in several mammalian species. Recently, RSA in humans during spatial navigation was reported, but systematic data regarding human REM sleep are lacking. Using mesio- temporal corticography with foramen ovale electrodes in epileptic patients, we report the presence of a 1.5–3-Hz synchronous rhythmic hippocampal oscillation seemingly specific to REM sleep. This oscillation is continuous during whole REM periods, is clearly observable by visual inspection, and appears in tonic and phasic REM sleep episodes equally. Quantitative analysis proved that this 1.5–3-Hz frequency band significantly differentiates REM sleep from waking and slow-wake sleep (SWS). No other frequency band proved to be significant or showed this high rhythmicity. Even in temporo-lateral surface recordings, although visually much less striking, the relative power of the 1.5–3-Hz frequency band differentiates REM sleep from other states with statistical significance. This could mean that the 1.5–3-Hz hippocampal RSA spreads over other cortical areas in humans as in other mammals. We suggest that this oscillation is the counterpart of the hippocampal theta of mammalian REM sleep, and that the 1.5–3-Hz delta EEG activity is a basic neurophysiological feature of human REM sleep.

KEY WORDS: mesio-temporal corticography; foramen ovale electrodes; hippocampal theta; rhythmic delta; temporal epilepsy

INTRODUCTION

Hippocampal rhythmic slow activity (RSA) appears during exploratory behavior and rapid eye movement (REM) sleep of rodents, cats, dogs, and other mammals (Robinson, 1980; Grastya ´n and Karmos, 1961;Winson, 1974; Vanderwolf, 1969; Lerma and Garcia-Ausst, 1985; O’Keefe, 1993; Arnolds et al., 1979; Brown, 1968). It is the most studied physiological marker of waking and REM sleep-related arousal/cognition in animal models, and its functional role in memory processing and spatial orientation is being increasingly clarified at the cellular level (Poe et al., 2000; Buzsáki et al., 1990; Buzsáki, 1996; Vinogradova, 1995; Haas et al., 1995; Greenstein et al., 1988; Winson, 1978, 1993).

Many hypotheses of REM sleep function and dreaming are based on hippocampal RSA (Poe et al., 2000; Winson, 1993; Sandyk, 1998). According to several theories, hippocampal RSA provides a mechanism for the consolidation of newly acquired memories and for weakening older ones in REM sleep (Poe et al., 2000;Winson, 1993). There are some hints that similar hippocampal RSA exists in primates, including humans (Robinson, 1980; Brazier, 1968; Wieser, 1984; Stewart and Fox, 1991; Mann et al., 1997). Recently, task-specific RSA was outlined by intracranial EEG recordings during computer-generated maze-navigation sessions in humans (Kahana et al., 1999), but there are no relevant data regarding human REM sleep (Robinson, 1980; Brazier, 1968; Wieser, 1984; Mann et al., 1997), and the results of animal experiments cannot be automatically transferred to humans (Robinson, 1980). Here we report our results on REM sleep hippocampal activity detected by foramen ovale (FO) electrodes in 12 epileptic patients undergoing presurgical examination.

METHODS

Patients (n = 12, 3 males, 9 females, between 21–61 years old) required mesio-temporal corticography to verify the location of a temporal lobe epileptogenic region prior to surgical resection. FO electrodes were introduced during general anesthesia. All patients gave written informed consent for the implantation of the electrodes and for participation in the study.

FO electrodes are flexible wires introduced through the foramen ovale into the cisterna ambiens, and they contact the parahippocampal gyrus at four points bilaterally along the axis of the hippocampal formation (Fig. 1). This kind of mesio-temporal corticography provides a unique opportunity to measure hippocampal population activity in a semi-invasive way, without opening the skull. FO electrodes contained four silver contacting points, each 5 mm apart. Ten out of 12 recordings were bilateral. Additional electrodes of the 10–20 system and standard polysomnography were used to define sleep stages according to standard criteria (Re- chtschaffen and Kales, 1968). Parallel scalp recordings of the 10–20 system, electro-oculography, submental electromyography, and electrocardiography were performed with Ag/AgCl electrodes fixed with collodium. X-rays of the skull determined the location of the FO electrodes. Signals from all electrodes were cutoff filtered at 0.33 Hz, and amp
lified and digitized at 128 Hz with 12-bit resolution.

Depending on the clinical indication, patients spent 3–12 days in the monitoring room, and were allowed to be sitting or lying. The patients were able to talk with visitors, watch TV, read, etc. The timing of lights off was determined by the patients at will. Continuous video-EEG monitoring was performed.

A 5–10-min recording made with the patient awake and with eyes open (without speaking and any major movement or watching TV) was stored for the waking-eyes open condition. Every night’s recording from lights off to spontaneous awakening was also stored on removable disks of 100-MB capacity. All other states were selected off-line from this stored data. We excluded the first night from our analysis and then selected representative samples of 4-s epochs for every patient’s every sleep-waking state, avoiding artifacts and epileptic spikes in all 32 simultaneously recorded channels. This was made visually according to the criteria of Rechtschaffen and Kales (1968). An epoch was accepted for quantitative analysis only when all 32 simultaneously recorded channels were free of artifacts and epileptic spikes. Thirty to 100 epochs were selected for every patient’s every sleep-waking state, which belonged to the recordings of night 2 or 3. The sleep-waking states defined here are the following: waking with eyes open (W-eo), waking with eyes closed (W-ec), light slow-wave sleep (LSWS) = stage 2 sleep, deep slow-wave sleep (DSWS) = stage 3 + 4 sleep, REM-To = tonic REM periods (REM sleep periods without eye movements), and REM-Ph = phasic REM periods (REM sleep periods accompanied by ocular motility). The standard reference point used in all patients was the vertex (Cz electrode). In those patients in whom contralateral mastoid and/or earlobe reference points were also present, the FO signals were analyzed referring them to these points, too.

A Hanning window was applied to all selected epochs, after which they were subjected to the fast Fourier transformation algorithm, resulting in periodograms with 0.25-Hz spectral resolution. An average was then calculated for each sleep-waking state and each FO and T3, T4 electrode of each subject. Relative power of the low delta (0–1.25 Hz), mid-delta (1.5–3 Hz), high delta (3.25–4.5 Hz), low theta (4.75–6.25 Hz), mid-theta (6.5–7.75 Hz), and high theta or low alpha (8–9.5Hz) bands was calculated.

A repeated measure one-way ANOVA for the state effect was then performed, with every frequency band as dependent variable separately, for a randomly selected FO electrode of each patient. Planned comparisons by means of a contrast analysis were also performed between the collapsed two conditions of every behavioral state (waking, slow-wave sleep, and REM sleep) and the collapsed data of other sleep-waking states (waking vs. slow-wave sleep + REM, slow-wave sleep vs. waking + REM, and REM vs. waking + slow-wave sleep). The same analysis with log-normalized relative power (log(R/1 – R); R + relative power), and with all FO electrodes (8/patient, 4 at each side) as cases, produced similar results, so the one electrode/subject-not normalized data are reported here.

RESULTS

FO recordings have the same general pattern of EEG activity as temporo-lateral scalp recordings except for REM sleep. Here, a striking 1.5–3-Hz continuous, rhythmic oscillation appears (Fig 2), while on the scalp the well-known EEG desynchronization is seen. Average spectral power based on fast Fourier transformation showed a consistent peak in the 1.5–3-Hz frequency band in REM sleep in all FO electrodes. Slow activity with this frequency appeared on both sides, and the side of epileptic activity does not differ from the functionally unaltered hippocampus, even in cases of unambiguous unilateral temporal epilepsy (Fig. 3). The mean coherence in the 1.5–3-Hz band was 0.40, and its standard deviation was 0.27. The oscillation is highly synchronous between all four recording points; hence it is not detectable in bipolar montages (Fig. 4). In addition to the Cz vertex reference, used in clinical settings, contralateral mastoid reference (2 patients) and earlobe reference (3 patients) were used. In all cases the same pattern of data emerged, which rules out the possibility that this highly synchronous rhythm is produced in the reference electrode. Moreover, the vertex-referred signals were of much higher amplitude than the earlobe- or mastoid-referred ones, suggesting that this RSA is produced near the FO electrodes. We used the vertex- referred signals in our statistical analysis.

Statistical comparisons of relative power values across different behavioral states (Fig. 5) show a strict correlation of 1.5–3-Hz activity with REM sleep, while low-delta power (0–1.25 Hz) was highest in slow-wave sleep. The two subdivisions of the delta range seemto have a reciprocal relation in REM and non-REM (NREM) sleep. Obviously there is an overlap between the delta frequency oscillation found in NREM and REM sleep, but relative power values create a sharper difference between the two sleep states. The main effect for state in the low delta band (0–1.25 Hz) is F(5, 45) = 29.32, P < 0.000001 (one-way ANOVA, repeated measures with six levels); contrast analysis between slow-wave sleep and all other states was highly significant (F(1, 9) = 44.22, P < 0.00009). The same analysis performed on relative power values of themid-delta band (1.5–3 Hz) yielded a significant effect for REM sleep, supported by themain effect for state (F(5, 45)= 11.54, P < 0.0000001) and by the contrast analysis between REM sleep and other sleep-waking states (F(1, 9) = 15.45, P < 0.003; see Fig.4). In waking, higher-frequency components are characteristic. Absolute power values are incomparable because of the extremely high power around 1 Hz of deep slow-wave sleep.

Although a clear association exists in some patients between the appearance of rapid eye movements and the 1.5–3-Hz oscillation, the difference between tonic and phasic REM periods does not reach the level of significance for any frequency band analyzed here, either for absolute or for relative power values.

In order to analyze the matter of spreading over other cortical areas of this hippocampal RSA in human REM sleep, we applied the same quantitative methods to signals of the T3/T4 temporal scalp electrodes as for the FO ones. Average spectral power did not show a consistent peak in the 1.5–3-Hz frequency band, but statistical analysis showed amain effect for state in the 1.5–3-Hz band (F(5, 45) = 10.63; P < 0.000001); the contrast analysis yielded REM sleep values significantly higher than the values for other states [F(1, 9) = 7.91; P < 0.02).

DISCUSSION

Rhythmicity is one of the most prominent properties of hippocampal RSA in animal experiments. In our study, one of the most striking feature of the 1.5–3-Hz activity detected with FO electrodes during human REM sleep was the high rhythmicity, which produced sharp peaks in the spectral curves (see Figs. 2, 3).

Synchrony or coherence are other major features of hippocampal RSA in different animal species (Robinson, 1980). This is similar to the case of the hippocampal RSA reported here, since it disappears and is practically undetectable in bipolarmontages (Fig. 4). The stability of the hippocampal RSA in different monopolar montages, and its increasing amplitude if the distance from the reference electrode increases, suggest that it is produced near the FO electrodes and is not an artifact or an activity pr
oduced in the reference electrodes.

As a sleep-dependent oscillation, the 1.5–3-Hz hippocampal RSA reported here seems to be specific for the REM phase. It does not appear in other sleep-waking states. Visual inspection of the raw signal and the statistical analysis of its spectral values support this specificity. However, the active behavioral and cognitively demanding situations which in other studies are known to be accompanied by hippocampal RSA were not analyzed. The 1.5–3-Hz human hippocampal RSA was continuous during an REM period. It did not appear in bursts like the RSA detected by intracranial electrodes during virtual maze navigation (Kahana et al., 1999), and it was practically uninterrupted by other EEG activities. This phenomenon recurred in all REM periods and appeared repeatedly during recording nights. This fact strongly supports the ideas that it is specific for REM sleep and is a tonic feature of this sleep state.

In rats, hippocampal RSA spreads over other cortical regions outside the hippocampus (Gerbrandt et al., 1978). In our human recordings we also found some evidence for this, i.e., spectral values of the 1.5–3-Hz frequency band showed some specificity for REM sleep beyond the cortical desynchronization present during this sleep state in the temporo-lateral leads. Further studies are needed to clarify the neocortical distribution of this frequency band during REM sleep in humans.

The REM sleep-dependent hippocampal RSA reported here is of lower frequency than those observed in animal studies, falling in the delta and not in the theta band. It must be mentioned that neither of the human studies reporting theta oscillation during various cognitively demanding situations proved unambiguously the hippocampal origin of this oscillation. Scalp recordings are unable to detect hippocampal population activity, and the intracranial recordings of Kahana et al. (1999) did not exhibit enhanced theta power near the hippocampal formation, as they state in their study. Depth hippocampal recordings during human REM sleep exhibited more delta than theta activity (Mann et al., 1997; Yu et al., 1997). Mann et al. (1997) interpreted this rhythmic delta as a theta analogue phenomenon in humans, whereas Yu et al. (1997) did not take into account the greater delta than theta power values. The latter argued for a relative increase in theta power during REM sleep as compared to stage 2 sleep. This REM sleep-specific hippocampal RSA can be clearly distinguished from the EEG delta activity of NREM sleep by its specific frequency range and its hippocampo-neocortical distribution (see our statistical analysis and Figs. 2 and 5). The lower frequency of REM sleep-specific hippocampal RSA in humans as compared to rats is similar to the finding of Bragin et al. (1999). They showed that hippocampal ripples are of 80–160 Hz in humans, whereas the frequency of ripples in rats is ca. 200Hz. This couldmean a general tendency of larger hippocampal formations producing slower rhythms.

Animal studies have shown that hippocampal RSA is of higher frequency during phasic REM periods than during tonic ones in rabbits (Harper, 1971), cats (Sakai et al., 1973), and rats (Robinson et al., 1977). This difference does not appear in our analysis of human REM sleep-related hippocampal RSA. Obviously further studies are needed to clarify the causes of this difference.

This rhythmic hippocampal slow activity could not be the consequence of the epileptic process, because the analyzed epochs are free from interictal epileptic discharges in all 32 simultaneously recorded channels. Even in epochs contaminated with spikes, the slow oscillation remained the same. Pathological EEG slowing did not contaminate REM sleep activity, even in the recordings of sides corresponding to seizure onset. Subclinical ictal activity could also be excluded, since the slow oscillation was observed continuously during the whole REM periods.

According to our results, rhythmic hippocampal slow oscillation does occur during REM sleep in humans, as found in other mammals. The rhythmicity, synchrony, state specificity, and spreading over other cortical areas are comparable to those of rodents’ REM sleep-related theta. It also seems to be a tonic feature of REM sleep. However, it is of lower frequency, falling in the delta and not in the theta band. The functional and comparative neurological significance of this frequency difference remains to be elucidated.

Most quantitative EEG studies of sleep are based on the misleading assumption that delta activity (usually defined as power in the 0.75–4.5-Hz frequency band) is a unique characteristic of NREM sleep, and they do not interpret the origin and significance of delta power in REM sleep (Borbély, 1982; Borbély and Achermann, 1999; Buchsbaum et al., 1982). A recent study based on principal component analysis of human sleep EEG concluded that the traditional division of the theta band in the human cortical EEG is artificial (Corsi-Cabrera et al., 2000). Our results suggest that a well-defined range of delta activity (1.5–3 Hz) in temporo-lateral leads is a basic characteristic of REMsleep, but has different sources from that of NREM delta oscillation. The latter results from the thalamo-cortical system, while the former is the result of hippocampo-cortical interplay.

Acknowledgments

The authors thank Dr. György Karmos for helpful discussions during the course of this research. 
Grant sponsor: Hungarian Medical Research Council; Grant number: 6090/7/ETT/2000

REFERENCES

Arnolds DE, Lopes da Silva FH, Aitink JW, Kamp A. 1979. Hippocampal EEG and behaviour in dog. I. Hippocampal EEG correlates of gross motor behaviour. Electroencephalogr Clin Neurophysiol 46:552–570.

Borbély AA. 1982. A two process model of sleep regulation. Hum Neurobiol 1:195–204.

Borbély AA, Achermann P. 1999. Sleep homeostasis and models of sleep regulation. J Biol Rhythms 14:557–568.

Bragin A, Engel J Jr, Wilson CL, Fried I, Buzsáki G. 1999. High-frequency oscillation in human brain. Hippocampus 9:137–142.

Brazier MA. 1968. Studies of the EEG activity of limbic structures inman. Electroencephalogr Clin Neurophysiol 25:309–318.

Brown BB. 1968. Frequency and phase of hippocampal theta activity in the spontaneously behaving cat. Electroencephalogr Clin Neurophysiol 24:53–62.

Buchsbaum MS, Mendelson WB, Duncan WC, Coppola R, Kelsoe J, Gillin JC. 1982. Topographic cortical mapping of EEG sleep stages during daytime naps in normal subjects. Sleep 5:248–255.

Buzsáki G. 1996. The hippocampo-neocortical dialogue. Cereb Cortex 6:81–92.

Buzsáki G, Chen LS, Gage FH. 1990. Spatial organization of physiological activity in the hippocampal region: relevance to memory formation. Prog Brain Res 83:257–268.

Corsi-Cabrera M, Guevara MA, Del Rio-Portilla Y, Arce C, Villanueva- Hernandez Y. 2000. EEG bands during wakefulness, slow-wave and paradoxical sleep as a result of principal component analysis in man. Sleep 23:738–744.

Gerbrandt LK, Lawrence JC, Eckardt MJ, Lloyd RL. 1978. Origin of the neocortically monitored theta rhythm in the curarized rat. Electroencephalogr Clin Neurophysiol 45:454–467.

Grastyán E, Karmos G. 1961. A study of a possible “dreaming” mechanism in the cat. Acta Physiol Acad Sci Hung 20:41–50.

Greenstein YJ, PavlidesC,Winson J. 1988. Long-term potentiation in the dentate gyrus is preferentially induced at theta rhythm periodicity. Brain Res 438:331–334.

Haas HL, Sergueeva OA, Vorobjev VS, Sharonova IN. 1995. Subcortical modulation of synaptic plasticity in the hippocampus. Behav Brain Res 66:41–44.

Harper RM. 1971. Frequency changes in hippocampal electrical activity during movement and tonic immobility. Physiol Behav 7:55–58.

Kahana MJ, Sekuler R, Caplan JB, Kirschen M, Mads
en JR. 1999. Human theta oscillation exhibits task dependence during virtual maze navigation. Nature 399:781–784.

Lerma J, Garcia-Ausst E. 1985. Hippocampal theta rhythm during paradoxical sleep. Effects of afferent stimuli and phase relationships with phasic events. Electroencephalogr Clin Neurophysiol 60:46–54.

Mann C, Simmons J,Wilson C, Engel J, Bragin A. 1997. EEG in human hippocampus, amygdala and entorhinal cortex during REM and NREM sleep. Sleep Res 26:27.

O’Keefe J. 1993. Hippocampus, theta, and spatial memory. Curr Opin Neurobiol 3:917–924.

Poe GR, Nitz DA, McNaughton BL, Barnes CA. 2000. Experience-dependent phase-reversal of hippocampal neuron firing during REM sleep. Brain Res 855:176–180.

Rechtschaffen A, Kales A. 1968. A manual of standardized terminology, techniques and scoring system for sleep stages of human subjects. Los Angeles: UCLA Brain Information Service.

Robinson TE. 1980. Hippocampal rhythmic slow activity (RSA; theta): a critical analysis of selected studies and discussion of possible species- differences. Brain Res 203:69–101.

Robinson TE, Kramis RC, Vanderwolf CH. 1977. Two types of cerebral activation during active sleep: relations to behavior. Brain Res 124: 544–549.

Sakai K, Sano K, Iwahara S. 1973. Eyemovements and hippocampal theta activity in cats. Electroencephalogr Clin Neurophysiol 34:547–549. Sandyk R. 1998. A neuromagnetic view of hippocampal memory functions. Int J Neurosci 93:251–256.

Stewart M, Fox SE. 1991. Hippocampal theta activity in monkeys. Brain Res 538:59–63.

Vanderwolf CH. 1969. Hippocampal electrical activity and voluntary movement in the rat. Electroencephalogr Clin Neurophysiol 26:407– 418.

Vinogradova OS. 1995. Expression, control, and probable functional significance of the neuronal theta-rhythm. Prog Neurobiol 45:523–583.

WieserHG. 1984. Temporal lobe epilepsy, sleep and arousal: stereo-EEG findings. In: Degen R, Niedermeyer E, editors. Epilepsy, sleep and sleep deprivation. Amsterdam: Elsevier Science Publishers B.V. p 137– 167.

Winson J. 1974. Patterns of hippocampal theta rhythm in the freely moving rat. Electroencephalogr Clin Neurophysiol 36:291–301.

Winson J. 1978. Loss of hippocampal theta rhythm results in spatial memory deficit in the rat. Science 201:160–163.

Winson J. 1993. The biology and function of rapid eye movement sleep. Curr Opin Neurobiol 3:243–248.

Yu Q, Rayport M, Farison JB, Dennis MJ, Choi YS. 1997. Computer analysis of human depth EEG in different sleep stages. Biomed Sci Instrum 33:184–190.