EEG spectral activity during paradoxical sleep

NEUROREPORT

SLEEP

EEG spectral activity during paradoxical sleep: further evidence for cognitive processing Christophe Jouny, Florian Chapotot1 and Helli MericaCA Unite de Neurophysiologie Clinique, Division de Neuropsychiatrie, HoÃpitaux Universitaires de GeneÁve, 2 Chemin du Petit BelAir, 1225 CheÃne-Bourg, GeneÁve, Switzerland; 1 Unite de Physiologie de la Vigilance, CRSSA Emile PardeÂ, La Tronche, France CA

Corresponding Author

Received 1 August 2000; accepted 6 September 2000

Paradoxical sleep (PS), in which periods with (phasic) and without (tonic) rapid eye movements are intermingled, is hypothesized to be related to cognitive processing and dreaming. Based on polysomnographic data from 12 healthy subjects, this study focuses on the spectral differentiation between phasic and tonic periods. Phasic PS periods exhibited decreased theta and alpha power in the posterior brain areas suggesting

the interference of visual processing related to dream imagery. Phasic PS periods were also characterized by a shift from beta to gamma activity in frontal, central and occipital areas re¯ecting speci®c phasic related activation. Together, these ®ndings bring new evidence for the existence of visual imagery and cognitive processing during phasic PS. NeuroReport 11:3667±3671 & 2000 Lippincott Williams & Wilkins.

Key words: Cognitive processing; Gamma activity, Paradoxical sleep; REM sleep; Spectral analysis

INTRODUCTION

Paradoxical sleep (PS) is characterized by the simultaneous presence of desynchronized EEG activity, atonia in the antigravity muscles and episodic periods of rapid eye movements (REMs). Periods with REMs are typically associated with muscle twitches and changes in breathing and heart rate [1] and constitute the phasic component of PS. These phasic periods are thought to play an important role in cognitive processing and dreams [2,3]. Several studies have demonstrated speci®c changes in behavioural and motor activity during phasic PS periods. For example, Kohyama and co-workers [4] showed a loss of EMG activity in mentalis and intercostal muscles during phasic PS that started before REM onset and persisted throughout the entire eye movement. Furthermore, phasic periods are characterized by a higher behavioural response threshold [5] and exhibit a lower auditory evoked response [6] than in tonic PS. These results support the hypothesis for a different involvement of areas in the generation of phasic [6±8] and tonic PS. They also suggest that the cognitive, motor and autonomic responses during phasic PS are re¯ected by activation of these areas producing speci®c changes in EEG activity. Considering the beta activity predominance in tonic PS [9] and the hypothesis that beta activity creates an EEG background for gamma synchronization during visual perception [10], it may be suggested that if cognitive processes and dreams occurred preferentially during phasic PS periods, then these processes could be accompanied by an increase in faster EEG frequencies, i.e. the gamma band. In fact, gamma activity has been

0959-4965 & Lippincott Williams & Wilkins

commonly associated with awareness and conscious perception [11] and described as an indicator of various cognitive processes as lexical processing [12]. It has also been demonstrated that gamma activity is not an alphaharmonic [13] but that it could manifest cognitive binding processes [14]. Most studies concerning EEG spectral activity in PS conducted so far have been restricted to the tonic PS periods because EEG signals may be contaminated by several sources during the phasic periods. For example, phasic events, such as middle-ear muscle activity (MEMA) [15], periorbital integrated potentials (PIPs) [16] and phasic motor inhibition potentials [4] overlap with EEG spectral density measures. Eye movements also contaminate lowfrequency EEG measures in frontal and central brain areas, while their in¯uence is negligible in the occipital area [9]. Using an ocular artefact removal procedure, Waterman et al. [17] found a decrease in alpha and beta power in central and occipital regions during phasic PS periods compared with tonic PS periods. These changes were associated with a rise in theta power in the frontal derivation. The decrease in centro-occipital alpha and beta power has been interpreted as a consequence of information processing occurring during phasic PS by analogy with the EEG response after stimulation in wakefulness [18]. The major limitation of that study is, however, that the analysis was restricted to EEG frequencies , 30 Hz, discarding faster frequencies which are known to be associated with cognitive tasks [19], visual processing such as detection of moving stimuli [20,21] and phasic PS sleep in epileptic patients [19].

Vol 11 No 17 27 November 2000

3667

NEUROREPORT

C. JOUNY, F. CHAPOTOT AND H. MERICA

Artefact and REM detection: EEG signals were high- and low-pass ®ltered (EEG 0.5±70 Hz, EOG 0.1±15 Hz, EMG 5± 70 Hz). The analogue signals were digitized at a sampling rate of 256 Hz with a 12-bit resolution and then stored on CD-ROM. Prior to analyses, the signals were subjected to an automatic 1 s resolution artefact detection routine using a background-dependent ®lter based on the root mean square amplitude of the signals. After visual validation, all 1 s epochs containing artefacts were coded as missing data. The automatic REM detection was obtained by an adaptation of the Takahashi and Atsumi algorithm [25]. The EOG was smoothed using an 80 ms sliding window and then differentiated twice to obtain respectively the slope and the two in¯exion points representing the start and the end of the eye movement. Three ®xed thresholds were used for the automatic detection: amplitude . 30 ìV, slope . 500 ìV/s and duration , 0.5 s. The automatic detection was visually validated.

In the present study, we investigated whether particular changes in EEG activities, especially in high-frequency bands, occur during phasic PS sleep but not during tonic PS, thus sustaining the visual and cognitive processing hypothesis. The spectral constituents in the EEG of tonic and phasic periods are compared using spectral analysis over a wide spectral range (0.5±45 Hz).

MATERIALS AND METHODS

Subjects: This is a retrospective study based on all-night sleep polygraphic data obtained from 12 healthy paid volunteers (three men and nine women) aged between 21 and 29 (mean 25.3  2.5) years who participated in studies on sleep structure. Informed written consent of each subject was obtained in accordance with Ethical Committee requirements. All subjects were screened for good health on the basis of their history and clinical examination and asked to refrain from excessive caffeine or alcohol consumption on the days preceding nocturnal polysomnography. Sleep recordings obtained in control conditions only were included in the analysis.

EEG spectral analysis of phasic vs tonic periods: EEG power spectra were computed for 2 s epochs using fast Fourier transform (FFT) algorithm with a Hanning window. Power spectra were then divided into six frequency bands: low delta (0±2 Hz), delta (2±5 Hz), theta (5±8 Hz), alpha (8±12 Hz), beta (12±35 Hz) and gamma (35±45 Hz) by summing power in 0.5 Hz frequency bins. For each band, the lower bin limit was excluded and the upper bin limit included. As no speci®c sigma activity could be observed during the PS episodes, the sigma band was not considered separately. Instead a wide beta band was used. Phasic and tonic PS periods were separated on the basis of the presence or absence of REMs in the 2 s FFT windows. To discard any border effect, only strings of three similar consecutive 2 s epochs, i.e. only epochs surrounded by an epoch of the same type on both sides, were taken to de®ne phasic and tonic periods. Since not all subjects provided strings during the ®rst PS episode, the phasic vs tonic comparison was undertaken only for the second, third and fourth PS episodes. An average of 133  29 phasic strings

EEG recordings and sleep scoring: All-night sleep was recorded using three bipolar EEG derivations (F4-CZ, C4T4, and PZ-O2), one horizontal electrooculogram (EOG), one submental electromyogram (EMG) and an ECG. Lights-off and lights-on time was scheduled at 22:00 h and 06:30 h. Sleep stages were visually scored in 20 s epochs by an experienced technician using Rechtschaffen and Kales rules [22]. PS episodes were determined using the 15 min combining rule [23,24] with no minimal duration required to de®ne the start of a PS episode. Calculation of sleep parameters included total sleep time (TST), sleep ef®ciency (SE ˆ TST/total recording time 3 100) and percentage of different sleep stages. PS duration, latency and stability were also calculated (de®nitions given in Table 1). All intruding intervals of other sleep stages or of wakefulness were considered as missing data so as to preserve PS time continuity.

Table 1. Paradoxical sleep parameters. Subject

Age (years)

1 2 3 4 5 6 7 8 9 10 11 12

24 23 26 26 27 21 27 29 26 27 21 27

Mean s.e.m.

25.3 0.9

PS latency (min)

PS episode duration (min)

PS stability (%)

Percentage of phasic periods

Number of phasic strings used

1

2

3

4

Total

74 141 100 89 70 82 108 65 67 65 107 139

14 7 7 13 14 9 15 18 13 5 11 30

37 20 41 20 43 22 17 23 28 24 15 11

50 32 39 29 39 30 28 50 36 18 18 40

38 32 40 38 47 56 42 51 24 30 73 27

147 117 127 124 191 118 102 148 132 132 152 132

73 85 68 93 82 93 83 89 89 73 88 81

28.9 22.3 25.5 24.7 35.7 25.0 23.4 27.5 26.4 24.8 29.7 26.2

70 37 40 129 176 35 145 410 79 130 155 195

92 8

13 2

25 3

34 3

41 4

135 6

83 2

26.7 1.0

133 29

PS latency is measured from sleep onset to the ®rst PS epoch; PS stability is the percentage of PS to PS transitions over all PS transitions within a PS episode.

3668

Vol 11 No 17 27 November 2000

NEUROREPORT

EEG CHARACTERISTICS OF PARADOXICAL SLEEP

was obtained from each subject, representing 20% of the total phasic period duration. To compensate for tonic predominance (73% of total PS duration) over phasic (27% of total PS duration), the same number of tonic strings as available phasic ones was randomly chosen across all valid tonic strings from each subject. Since spectral changes in high frequencies may be related to muscle contamination we performed a similar spectral analysis on the chin EMG as for the EEG channels. To achieve comparison between phasic and tonic PS periods, spectral EMG data were taken from periods corresponding to the 2 s strings used for EEG spectral comparison. All recordings were analysed using the ERA software package (Phitools, Grenoble, France) for sleep stage scoring and quantitative analyses. Statistical analysis: A three-way ANOVA for repeated measures with factors derivation (F4-CZ, C4-T4 and PZO2), PS episode (2±4) and condition (tonic/phasic) was performed on the EEG spectral power in each frequency band. When signi®cant effects were present ( p < 0.05) posthoc differences between conditions were evaluated using the non-parametric Wilcoxon paired test. Data are expressed as mean  s.e.m.

RESULTS

Mean spectral power (µV2/Hz)

Sleep variables: Table 1 gives the polygraphic characteristics for the subjects studied. Standard sleep variables were in the normal range with a mean total sleep time of 506  7 min and a mean sleep ef®ciency of 93.1  1.1%. As expected, PS episode duration gradually increased with

400

Theta 25

100

***

80

300

***

*

*

20

60

15

40

10

20

5

200 100 0

F4–Cz C4–T4 Pz–O2

0

F4–Cz C4–T4 Pz–O2

Alpha Mean spectral power (µV2/Hz)

Phasic versus tonic EEG spectral activities: Figure 1 shows the mean power at each derivation during phasic and tonic PS periods. A signi®cant interaction between condition and derivation was found for all frequency bands except the beta band for which a signi®cant condition effect was present without interaction. As no signi®cant cycle effect was found, EEG powers in the three cycles were averaged for post-hoc tests. Comparison of tonic and phasic periods showed a different pattern for each considered frequency band. As could be expected, the slow frequency bands increased signi®cantly at the frontal derivation during phasic PS periods, while posterior derivation remained unchanged. No signi®cant difference was seen at the occipital lead for either the low delta or the delta frequency bands. A similar anterior increase was present for theta power even though the differences did not reach statistical signi®cance. Interestingly, although the theta level did not differ across the scalp during tonic periods, the phasic PS periods were characterized by a signi®cant decrease at the PZ-O2 lead. This decrease in theta power was accompanied by a concomitant decrease in alpha power in the posterior area. The beta frequency band did not show differences between electrodes during the tonic periods. Phasic PS periods were characterized by a signi®cant decrease in beta

Delta

Low-delta ***

time across the night, with a concomitant increase in the number of phasic periods. Thus, the percentage of phasic vs tonic periods across the three episodes may be considered constant.

F4–Cz C4–T4 Pz–O2

Gamma

Beta 1.5

15

30 25

0

**

*

20

**

**

***

10

1

5

0.5

*

**

15 10 5 0

F4–Cz C4–T4 Pz–O2

0

0 F4–Cz C4–T4 Pz–O2

F4–Cz C4–T4 Pz–O2

Fig. 1. Phasic vs tonic comparison in the mean power levels for the different spectral bands at the three derivations. Vertical lines indicate s.e.m. White bars represent the tonic condition and grey bars the phasic condition. Signi®cant difference between conditions:  p < 0.05;  p < 0.01;  p < 0.001.

Vol 11 No 17 27 November 2000

3669

NEUROREPORT

C. JOUNY, F. CHAPOTOT AND H. MERICA

power of approximately equal magnitude at all considered derivations. Unlike beta, the gamma frequency band increased during phasic PS periods in all areas, with higher amplitude at the frontal lead. Figure 2 shows the percentage difference in spectral power between phasic and tonic periods for the central and posterior EEG derivations and for the EMG. Positive values indicate an increase and negative values a decrease in power during phasic periods. Signi®cant increase in EEG gamma power was noted during the phasic PS periods as described above. Simultaneous analysis of the EMG derivation revealed that above 15 Hz a signi®cant decrease in EMG activity was present, indicating the absence of a myographic contamination effect.

DISCUSSION

Phasic vs tonic spectral power difference (%)

The aim of this study was to analyse EEG activity during PS to characterize EEG changes during phasic and tonic PS periods and, in particular, to assess the characteristics of high frequency EEG activities during phasic periods to elucidate possible speci®c processes. In agreement with a previous study [9], phasic PS periods were characterized by a signi®cant decrease in the theta, alpha and beta activities. Our study shows, however, that the observed decrease in the beta band power was associated with a signi®cant rise in gamma power suggesting that processes in¯uencing phasic PS periods are essentially expressed by changes in the faster (.35 Hz) frequencies. The difference in high-frequency EEG activities supports the hypothesis for the existence of different activation processes involved in phasic vs tonic PS. Our results show that signi®cant EEG changes are detected during phasic PS. While an increase in the low EEG frequencies in anterior and central areas may be easily related to eye movements, changes in theta activity are more dif®cult to interpret. Although a portion of the increased theta activity during phasic PS periods in the frontal and central regions (Fig. 1) may be attributed to eye

50 C4–T4 25

0

Pz–O2 EMG

225

250

2

5

8

12 15

25 Frequency (Hz)

35

45

Fig. 2. Phasic vs tonic mean power spectrum difference expressed as percentage of tonic mean level at two EEG derivations (closed circles, C4-T4; open circles, PZ-O2) and at the chin EMG derivation. Vertical dotted lines indicate spectral limits used to de®ne frequency bands. F4CZ is not represented to permit visualisation of small differences.

3670

Vol 11 No 17 27 November 2000

movement contamination, the decrease in the posterior area associated with a decrease in alpha activity suggests a pattern of EEG activation speci®c to cognitive and memory processes related to visual imagery and dreaming [26]. By analogy with waking EEG, decreased occipital alpha activity during the phasic PS periods may be viewed as an expression of visual cortex activation related to information processing during visual imagery [9] and as an alphablocking reaction to complex mental imagery [27]. The main ®nding of our study comes from the detailed analysis of the low-voltage high frequency EEG that typically characterizes PS. Our results con®rm previous reports indicating decreased beta activity during these periods compared to tonic PS periods in all regions examined. Waterman [9] and Uchida [28] suggested that the decreased beta power in the phasic PS periods could be related to an increased level of arousal resulting from the REM activity [29] that desynchronizes the beta activity. In the light of our data, however, another hypothesis can be proposed. In accordance with our initial supposition, phasic periods were associated with an increased activity in gamma frequency band. Phasic PS periods induced a shift in high frequencies from beta to gamma band. So that the decrease in beta activity related to phasic periods is not due to an EEG desynchronization but in fact to a shift to faster EEG spectral activities. In support of this hypothesis Gross and Gotman [19] showed that, using intracranial ECoG measurements in a population of epileptic patients, gamma activity increased during a cognitive task with the same pattern as that obtained during phasic PS periods. The phasic increase in the gamma band was associated with a reduction in beta activity similar to that found in our healthy subjects. Thus the differences in beta and gamma activity patterns during phasic and tonic PS may well re¯ect a different level of activation related to different processes of tonic and phasic PS. Evoked potential studies have shown that different high frequency oscillatory activity (beta and gamma) may represent different aspects of visual processing [30]. We cannot con®dently af®rm that the gamma band power measurements are measurements of electrical activity of the brain alone, so this interpretation is given with some reserve. The EMG analysis was done on the mentalis muscle, which does not exclude muscular contamination related to other phasic events such as PIPs and MEMAs. We could nevertheless argue that since an inverse relationship was found between EMG spectral activity and gamma power, we do not believe that the gamma increase may be merely related to facial muscular contamination. Moreover, any muscle contamination would affect the entire frequency range, which does not ®t with the observed decrease in beta activity. The same argument holds for direct EEG contamination by the abrupt deviation of the eye movement per se. Nevertheless, part of the increased gamma activity in frontal areas only could still be due to contamination by PIPs.

CONCLUSION

Our study brings out speci®c changes in EEG activities during phasic PS periods consisting in an increase in gamma power, and a concomitant decrease in theta, alpha and beta activities. These speci®c phasic spectral changes

NEUROREPORT

EEG CHARACTERISTICS OF PARADOXICAL SLEEP

support the hypothesis of visual imagery and cognitive processes occurring preferentially during phasic PS periods. Further studies are needed to strengthen this hypothesis, involving sleep manipulation in an attempt to elucidate the interaction of different regulatory mechanisms involved in the production of phasic and tonic PS.

REFERENCES

1. Coccagna G, Mantovani M, Brignani F et al. Bull Physiopathol Resp 8, 1159±72 (1972). 2. Dement WC and Kleitman N. J Exp Psychol 53, 339±346 (1957). 3. Hobson JA and McCarley R. Am J Psychiatry 134, 1335±1348 (1977). 4. Kohyama J, Shimohira M, Hasegawa T et al. Exp Brain Res 107, 137±144 (1995). 5. Price LJ and Kremen I. Psychophysiology 17, 133±140 (1980). 6. Ornitz EM, Ritvo ER, Carr EM et al. Electroencephalogr Clin Neurophysiol 22, 514±524 (1967). 7. Hobson JA, Lydic R and Baghdoyan HA. Behav Brain Sci 9, 371±448 (1986). 8. Kohyama J, Tachibana N and Taniguchi M. Acta Neurol Scand 99, 368±373 (1999). 9. Waterman D, Elton M, Hofman W et al. J Sleep Res 2, 21±27 (1992). 10. WroÁbel A. Acta Neurobiol Exp 60, 247±260 (2000). 11. Jokeit H and Makeig S. Proc Natl Acad Sci USA 91, 6339±6343 (1994). 12. PulvermuÈller F, Lutzenberger W, Preissl H et al. Neuroreport 6, 2059±2064 (1995).

13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

23. 24. 25. 26. 27. 28. 29. 30.

Muller MM. Acta Neurobiol Exp (Warsz) 60, 49±65 (2000). Klemm WR, Li TH and Hernandez JL. Conscious Cogn 9, 66±85 (2000). Foulkes D and Pope R. Percept Motor Skill 37, 107±118 (1973). Rechtschaffen A. Conference of thinking. Hollins College, Roanoke, Virginia. 1971: 1±53. Waterman D, Woestenburg JC, Elton M et al. Sleep 15, 371±375 (1992). Davidson RJ, Chapman JP, Chapman LJ et al. Psychophysiology 27, 528±543 (1990). Gross DW and Gotman J. Neuroscience 94, 1005±1018 (1999). Gruber T, Muller MM, Keil A et al. Clin Neurophysiol 110, 2074±2085 (1999). Slobounov SM, Fukada K, Simon R et al. Brain Res Cogn Brain Res 9, 177±192 (2000). Rechtschaffen A and Kales A. A Manual of Standardized Terminology, Techniques and Scoring System for Sleep Stages of Human Subjects. Washington DC: Public Health Service, US Government Printing Of®ce; 1968. Feinberg I and Floyd TC. Psychophysiology 16, 283±291 (1979). Merica H and Gaillard JM. Physiol Behav 50, 1153±1159 (1991). Takahashi K and Atsumi Y. Sleep 20, 743±752 (1997). Klimesch W. Brain Res Rev 29, 169±195 (1999). Cantero JL, Atienza M, Salas RM et al. Neuroreport 10, 1817±1820 (1999). Uchida S, Maloney T and Feinberg I. Sleep 15, 352±358 (1992). Feinberg I, Floyd TC and March JD. Electroencephalogr Clin Neurophysiol 67, 217±221 (1987) Tzelepi A, Bezerianos T and Bodis-Wollner I. Clin Neurophysiol 111, 259±269 (2000).

Acknowledgements: We thank Dr Robert Blois for providing the data and Barbara Bertram and Ashok Lalji for their technical assistance. We are also grateful to Dr Alain Buguet for his support in the development of the ERA software. This research is supported by the Swiss National Science Foundation Grant No. 3100-050765.97.

Vol 11 No 17 27 November 2000

3671