Proc. Natl. Acad. Sci. USAVol. 93, pp. 2533-2538, March 1996Neurobiology

Intracortical and corticothalamic coherency of fastspontaneous oscillationsMIRCEA STERIADE* AND FLORIN AMZICALaboratoire de Neurophysiologie, Faculte de Medecine, Universite Laval, QC Canada GlK 7P4

Communicated by R. Llinds, New York University Medical Center, New York, NY, October 25, 1995

ABSTRACT We report that fast (mainly 30- to 40-Hz)coherent electric field oscillations appear spontaneously dur-ing brain activation, as expressed by electroencephalogram(EEG) rhythlms, and they outlast the stimulation of mesopon-tine cholinergic nuclei in acutely prepared cats. The fastoscillations also appear during the sleep-like EEG patterns ofketamine/xylazine anesthesia, but they are selectively sup-pressed during the prolonged phase of the slow (< 1-Hz) sleeposcillation that is associated with hyperpolarization of corti-cal neurons. The fast (30- to 40-Hz) rhythms are synchronizedintracortically within vertical columns, among closely locatedcortical foci, and through reciprocal corticothalamic net-works. The fast oscillations do not reverse throughout thedepth of the cortex. This aspect stands in contrast with theconventional depth profile of evoked potentials and slow sleeposcillations that display opposite polarity at the surface andmidlayers. Current-source-density analyses reveal that thefast oscillations are associated with alternating microsinksand microsources across the cortex, while the evoked poten-tials and the slow oscillation display a massive current sink inmidlayers, confined by two sources in superficial and deeplayers. The synchronization of fast rhythms and their highamplitudes indicate that the term "EEG desynchronization,"used to designate brain-aroused states, is incorrect andshould be replaced with the original term, "EEG activation"[Moruzzi, G. & Magoun, H. W. (1949) Electroencephalogr. Clin.Neurophysiol. 1, 455-473].

Fast cortical rhythms, within a broad frequency range from 20to 80 Hz, are elicited by sensory stimuli or appear under avariety of behavioral conditions reflecting an increased alert-ness, such as focused attention, tasks requiring complex sen-sorimotor integration, and conditioned reflexes (reviewed inref. 1). Fast rhythms also occur spontaneously, as part of thebackground electrical activity of the brain, during both brain-active behavioral states of waking and rapid-eye-movement(REM) sleep in humans (2-4) and animals (5). These oscil-lations outlast the period of stimulation ofbrainstem ascendingcircuits that produce generalized activating effects on thalamo-cortical systems (6, 7). The mechanisms underlying the fastoscillations include the intrinsic properties of cortical (8) andthalamic (6, 9) neurons, together with synaptic couplingswithin intracortical and corticothalamic networks (5, 10).

In this paper, we focus on the intracortical and corticotha-lamic synchronization of fast oscillations evoked by stimulationof pedunculopontine tegmental (PPT) cholinergic nucleus orappearing spontaneously during activated epochs of the elec-troencephalogram (EEG). Our analyses demonstrate that thefast oscillation is distributed without phase reversal throughoutthe cortical depth, in contrast with the more conventionaldepth profile of the slow sleep oscillation or the early potentialevoked by PPT stimulation.

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

METHODSExperiments were conducted on adult cats anesthetized withketamine and xylazine (10-15 mg/kg and 2-3 mg/kg, respec-tively, i.m.). The animals were paralyzed with Flaxedil andartificially ventilated. End-tidal CO2 was maintained at 3.5-3.8% and body temperature at 37-39°C, heartbeat was mon-itored, and the EEG was continuously recorded to maintain aconstant sleep-like pattern by administering additional dosesof the anesthetic at the slightest signs of diminished amplitudesand/or increased frequencies of EEG rhythms.

Field potential recordings were performed from the supra-sylvian gyrus (areas 5 and 7) by means of (a) coaxial macro-electrodes, with the ring placed on the cortical surface and thetip inserted at a depth of 0.8-1 mm, and (b) circular arrays of6-8 electrodes separated by 0.25 mm, at successive depths. Theelectrodes were inserted perpendicularly to the surface andproduced monopolar recordings versus a reference placed onneck muscle. Besides, field potentials from area 5 and thethalamic intralaminar centrolateral (CL) nucleus were simul-taneously recorded. Intracellular recordings were performedin areas 5 and 7 with glass micropipettes filled with 3 Mpotassium acetate (impedance, 25-35 Mfl). A high-impedancedual amplifier with active bridge circuitry was used to recordand inject current into neurons. The signals were recorded onan eight-channel tape with a bandpass of 0-9 kHz, digitized at20 kHz for off-line computer analysis.The PPT nucleus was stimulated through bipolar electrodes

by using long pulse trains at 300 Hz or two to five short delayed(3-ms) pulses (see Figs. 1 and 2). The duration and intensity ofstimuli were 0.1-0.2 ms and 0.05-0.8 mA. Scopolamine wasadministered (0.5-1 mg/kg, i.v.) to block the PPT-elicitedprolonged depolarization of cortical neurons. Animals wereperfused with 10% formaldehyde under deep pentobarbitalanesthesia and the location of PPT stimulating electrodes wasexamined in frozen sections (80 ,um) stained with cresyl violetor thionine.We used auto- and cross-correlations (11). Three-dimen-

sional surfaces and topograms (contour maps) were con-structed from sequential cross-correlations (see Fig. 4; also seeref. 12). Current-source-density (CSD) analyses were alsoperformed on either evoked or spontaneous activities. Thecurrent flowing into or out of the membrane is proportional tothe second spatial derivative of the potential (13). The calcu-lation of this second derivative is made according to thefollowing formula (see ref. 14):

a24_) (z + n-Az) - 2+(z) + 4(z - nAz)az2 (n_AZ)2

where +A(z) is the potential at location z, Az is the distancebetween adjacent recording sites (in our case, Az = 0.25 mm,

Abbreviations: CL, centrolateral; CSD, current-source-density; EEG,electroencephalogram; EPSP, excitatory postsynaptic potential; PGO,ponto-geniculo-occipital wave; PPT, pedunculopontine tegmental nu-cleus; REM, rapid-eye-movement.*To whom reprint requests should be addressed.

2533

Dow

nloa

ded

by g

uest

on

Mar

ch 5

, 202

1

2534 Neurobiology: Steriade and Amzica

A

PPT 0.1 s

BI

B2 + B3

43 0.1is

the distance between two electrodes), and niz represents thedifferentiation grid (in our case, n = 1).

Topograms were generated by uniting with a line all pointswith the same height and by shading correspondingly the innerpart of the closed curve (see Fig. 1). The convention assigns

white to high positive values, while negative values are shadedwith black. Thus, sinks correspond to black and sources to white.

RESULTS

Stimulation of PPT nucleus suppressed the slow oscillation(<1 Hz) and produced fast rhythms between 20 and 60 Hz,mainly at 30-40 Hz. This effect, observed on gross recordings

FIG. 1. Depth profile of corti-cal fast oscillations induced by PPTstimulation, as revealed by sinksand sources. (A) Averaged traces(n = 15) of responses from area 5to two short-delayed PPT stimuli(arrow). The response started witha surface-positive (depth-negative)wave lasting for about 60 ms, fol-lowed by fast waves (about 35 Hz)that were in phase at the surfaceand at different depths. (B) CSDtopograms for the unfiltered re-sponse (BJ) and for the filteredresponse (15-80 Hz; B2) show athree-laminar distribution of thecurrents. Underlined sequence in40 Inls B2 is expanded at right (B3).

as well as by means of autocorrelation analyses, outlasted thestimulation period (5). Cross-correlations from simultaneousrecordings across areas 5 and 7 revealed that both slow and fastrhythms were coherent among various cortical foci, althoughthe slow oscillation was synchronized over wide distances (15),while the amplitude of correlation peaks of the fast rhythmsdecreased rapidly with the distance (see figure 12 in ref. 5).To analyze the depth profile of fast cortical oscillations, PPT

was stimulated with short pulse trains to elicit an earlypotential resembling the ponto-geniculo-occipital (PGO)waves ofREM sleep (16), followed by fast oscillations (n = 8).Thus, the depth profiles of the two events (early evokedpotential and subsequent fast rhythms) could be compared.

Proc. Natl. Acad. Sci. USA 93 (1996)

Dow

nloa

ded

by g

uest

on

Mar

ch 5

, 202

1

Proc. Natl. Acad. Sci. USA 93 (1996) 2535

A

1 I:

Control

scopol.B nn <

PPT

B

'A~~~~~~~4r

K,,Control

.......................................................................................................

I;'

PPT 0.2 s

FIG. 2. Prolonged cortical depolarization elicited by PPT stimulation is blocked by scopolamine. (A) Three superimposed responses of area 5,regular-spiking neuron (depth, 0.8 mm) to PPT pulse train (five stimuli at 300 Hz), before (Control) and after (Scopol.) systemic administrationof the muscarinic antagonist scopolamine. (B) Average of 15 PPT-evoked responses before and after scopolamine administration. Dotted linetentatively indicates the baseline. Membrane potential was - 75 mV before, and - 78 mV after, scopolamine administration. PPT stimulation elicitedan early EPSP (see expanded Inset in B) and a delayed, sustained depolarization (latency, about 150-200 ms), lasting for 2-2.5 sec. The lattercomponent was abolished by scopolamine.

The PPT-evoked early potential was biphasic, initially surface-positive, and reversed its polarity between 0.25 and 0.4 mm(Fig. 1A), as is also the case for spontaneously occurring PGOwaves during natural REM sleep in cats (unpublished data).Distinctly, the subsequent fast oscillations, lasting for 600-800ms, were in phase from the surface to the deepest corticallayers (Fig. IA). CSD topograms for the unfiltered activityevoked by the PPT pulse train showed a three-laminar distri-bution of currents, with an initial massive sink at depthscorresponding to layers 2-4, indicative for the PPT-thalamocortical input, mainly reaching midlayers (Fig. 1BJ).Indeed, mesopontine cholinergic nuclei have negligible, if any,

direct cortical projections and they reach the suprasylvianareas 5 and 7 through intermediate relays in thalamic associ-ation nuclei, the pulvinar-lateroposterior complex (17). Onthe other hand, the in-phase fast rhythms, analyzed in filteredactivity (between 15 and 80 Hz), displayed a pattern of sinksand sources that alternated at the frequency of about 35 Hz(Fig. 1 B2 and B3).The same type of extremely short pulse trains to the PPT

nucleus was used to obtain evidence of intracellular events thatdevelop during the period of fast rhythms in field potentialrecordings (n = 12). Following an early excitatory postsynapticpotential (EPSP) at a latency of 7 ms, consistent with its

... \. nl1

Neurobiology: Steriade and Amzica

'Ap.

Dow

nloa

ded

by g

uest

on

Mar

ch 5

, 202

1

2536 Neurobiology: Steriade and Amzica

AlSurface

0.25 mmm-

0.5 mmini -

0.75 mmm 1'

I mimn ^

1.25mm

1.5 mmi

~~~~~~- VfrP>--7 ,w,N~Sr Sodr1.75 mm I

2Sr l/ .--------

\r ~0.5$

A2

HELW)lo

BI

M4 liVvr\f I{n/v/\V\/%jtvj\x% l /1f/\yAQXuf\/\mrn/'-x 7v,wvfkJVWvrv\/\Y

A\,\tr -,QA

S l

3 40 ms

FIG. 3. Depth profile of spontaneously occurring slow and fast cortical oscillations. (A) Simultaneous recording through multiple electrodesinserted at various depths of area 5. At left (Al), averaged sweeps (n = 25) centered on the depth-negative sharp deflection of the slow oscillation.At right (A2), CSD analysis of these waves. (Bl1) Averages from the same sweeps as in Al (note different time scale), filtered between 15 and 80Hz, display in-phase fast (about 35 Hz) oscillations, transiently suppressed during the depth-positive phase of the slow oscillation. To allow a directcomparison between the unfiltered (Al) and filtered (B1) activities, an unfiltered wave from cortical depth inAl (thick trace) is superimposed onfiltered activities. (B2) CSD analysis of the period indicated by horizontal bar and arrow in Bl. Part marked by horizontal bar in B2 is expandedin B3. Note fast (about 35-Hz) alternating sinks and sources.

mediation through a bisynaptic PPT-thalamocortical pathway(see above), a sustained depolarization was observed, with alatency that varied between 100 and 600 ms in different cells,lasting for 2-3 s (Fig. 2). Systemic administration of themuscarinic blocker scopolamine abolished the sustained de-polarization while leaving intact the early EPSP (Fig. 2).CSD analyses are generally used for evoked potentials (14),

whereas they have only very recently been applied to hip-

pocampal spontaneous activities (18). We averaged the ste-reotyped waves building up the spontaneous slow oscillation(<1 Hz) of neocortex. This slow rhythm consists of a long-lasting depth positivity associated with hyperpolarization ofneurons, followed by a sharp negative field deflection associ-ated with depolarization of cortical neurons, eventually lead-ing to spindles and/or fast rhythms (19). Since the spikydepth-negative field potential reflects a massive and synchro-

IEI oHoIc

0.2 s

Proc. Natl. Acad. Sci. USA 93 (1996)

Dow

nloa

ded

by g

uest

on

Mar

ch 5

, 202

1

Proc. Natl. Acad. Sci. USA 93 (1996) 2537

CORTICO - THALAMIC10 _

8

'10

4

100 ' 2...

2

-100 0 100 ms

10:ACTIVATED

8

().5() .4

-0.5

1)

"s

10 6.

lO00 2

-100AVGarea 5

i\ ~/

CLCL

area 5

50 ms

nous excitatory event, its peak was chosen as time 0 for theselection of sweeps that were extracted as symmetrical win-dows (+1 s) around time 0. The same time stamps were alsoused when the digitally filtered data (usually 15-80 Hz) were

averaged (Fig. 3). The major components of the slow oscilla-tion were of opposite polarities at the surface and depth, beingreversed between 0.25 and 0.5 mm (Fig. 3A1). The correspond-ing CSD displayed a current sink in midlayers, confined by twosources and a less restricted sink in the deepest layer (Fig. 3A2).By contrast, the spontaneously occurring fast activities (35 Hz)were in phase from the surface down to 1.5-1.75 mm and were

selectively suppressed during the long-lasting depth positivityof the slow oscillation (Fig. 3B1). The CSD from fast filteredactivity showed alternating microsinks and microsources, re-

peated at the frequency (35 Hz) of the fast oscillation.To determine whether the intracortical synchronization of

the spontaneous fast oscillation (intracolumnar as well as

extending horizontally to adjacent cortical foci; see ref. 5) isaccompanied by coherent activities between cortical areas andrelated dorsal thalamic nuclei, we performed simultaneousrecordings from area 5 and intralaminar CL nucleus, known to

FIG. 4. Corticothalamic syn-chronization of fast rhythms. Se-quential cross-correlations(three-dimensional surfaces atleft, topograms at right; see tech-nical details in ref. 12) from pe-riods with slowly oscillating pat-terns characteristic for sleep(Top) and from an activated ep-och following PPT stimulation(Middle). Cross-correlationswere computed between filtered

(15- to 80-Hz) EEG waves re-

corded from the depth of cortical

area 5 and the thalamic intralami-

nar CL nucleus and derive from

sequential windows of 200 ms

_bd _ each. Note the increased cortico-O 100 ms thalamic synchrony during the

PPT-activated epoch and the co-

herence of the 25-Hz oscillation(see four symmetrical secondarypeaks at about 40 Hz). (Bottom)Averaged (n = 30) monosynapticresponses (latencies, 2-3 ms)evoked by CL stimulus in area 5and evoked by area 5 stimulus inCL nucleus, recorded throughthe same cortical and thalamicelectrodes that provided data inabove panels.

be reciprocally connected (ref. 20; see also Fig. 4 Bottom).Clear-cut evidence for corticothalamic synchronization of fastactivities was found in those instances in which the sites ofrecordings were identified to be reciprocally connected byusing monosynaptic corticothalamic and thalamocortical re-

sponses (n = 8). Fig. 4 illustrates such an experiment, showingthat fast spontaneous activities recorded from area 5 and CLnucleus were synchronized during sleep-like periods and thatduring the activated epochs elicited by PPT stimulation thecorticothalamic coherency increased, with four symmetricalsecondary peaks at about 40 ms (see the topogram derivedfrom sequential cross-correlations).

DISCUSSION

Two unexpected data are reported in the present paper. Thefirst is that, in an overwhelming majority of tracks, fastoscillations were found to be in phase throughout the corticaldepth. This might suggest volume conduction. However, thepresence of action potentials over the negative component offast field potentials recorded from superficial to deepest

SLEEP

(). ,

()4-0).5::

Neurobiology: Steriade and Amzica

Dow

nloa

ded

by g

uest

on

Mar

ch 5

, 202

1

2538 Neurobiology: Steriade and Amzica

cortical layers indicates that the fields are locally generated (5).In addition, a dramatic reduction of fast activity was observedin the underlying white matter, and short time lags weredetected between surface and depth activities (5). The CSDanalyses showed alternatively distributed microsinks and mi-crosources in both PPT-elicited and spontaneously occurringfast oscillations (Figs. 1 and 3). The smaller currents along thevertically organized dendritic core conductors, compared withtransmembrane ones, could account for the absence of poten-tial reversal of fast oscillations across the cortical depth. Thelimitation of the present study is that we used only six to eightrecorded foci, separated by 0.25 mm. Therefore, the numberof sinks and sources observed in Fig. 1 B2 and B3 and Fig. 3B2 and B3 may be higher with more and closer foci. What webasically demonstrate here is the difference between the depthreversal of the PPT-evoked potential or slow oscillation (witha massive current sink in midlayers, confined by two sources insuperficial and deep layers) and the absence of reversal of thefast oscillation that is associated with multiple sinks andsources throughout the depth of the cortex.The other surprising result is the presence of fast oscillations

during sleep-like activity patterns. This challenges the conven-tional view that high-frequency EEG waves are present onlyduring brain-activated states. However, distinctly from toni-cally activated patterns mimicking brain arousal, when fastoscillations appear in a sustained manner, during the slowsleep oscillation fast rhythms are- selectively suppressedthroughout the long-lasting depth-positive wave (Fig. 3) whichreflects prolonged hyperpolarization of neocortical neurons(19). Thus, besides their dependence on synaptic inputs withinintracortical and thalamocortical networks, fast rhythms arefavored by the depolarization of cortical cells and do not occurduring hyperpolarization. This voltage dependency of fastoscillations was demonstrated in local-circuit, sparsely spinousneurons (8) and in long-axoned, corticocortical and cortico-thalamic neurons (5, 21). Similarly, the depolarization depen-dency of fast oscillations was reported in a variety of thalamo-cortical neurons (6, 9). It is known that PPT stimulation, whicheffectively triggered fast oscillations in the present experi-ments, depolarizes and increases the apparent input resistanceof thalamocortical cells (22), a cholinergic effect mediated bymuscarinic receptors (refs. 22 and 23; also see Fig. 2). The PPTactivating cholinergic effect is transmitted to cortical neuronsthrough a bisynaptic, cholinergic-glutamatergic PPT-thala-mocortical pathway, as fast oscillations triggered by mesopon-tine cholinergic stimulation survive total lesions of the basalforebrain (6).We have also demonstrated that the intracortical coherency

of fast oscillations is coupled with synchronized fast rhythmsin corticothalamic circuits. As the synchronized oscillationswere revealed in reciprocal intralaminar-cortical projections(Fig. 4), these data support the hypothesis (3, 10) that thalamic

intralaminar nuclei, with widespread cortical connections (20)and possessing a special type of neurons discharging high-frequency rhythmic spike bursts at about 40 Hz during wakingandREM sleep (9), mediate the diffuseness of fast oscillations.Recent data in this laboratory revealed that in other systemstoo, including the visual and somatosensory systems, fastoscillations are robustly synchronized within corticothalamiccircuits during brain-activated states as well as during thedepolarizing phase of the slow sleep oscillation (24).

This research was supported by the Medical Research Council ofCanada (Grant MT-3689). F.A. is a Ph.D. student, partially supportedby the Fonds de Recherche en Sante du Quebec. We thank P. Giguereand D. Drolet for technical assistance.

1. Gray, C. M. (1994) J. Comput. Neurosci. 1, 11-38.2. Llinas, R. R. & Pare, D. (1991) Neuroscience 44, 521-535.3. Llinas, R. R. & Ribary, U. (1993) Proc. Natl. Acad. Sci. USA 90,

2078-2081.4. Uchida, S., Maloney, T. & Feinberg, I. (1994) Brain Res. 659,

243-248.5. Steriade, M., Amzica, F. & Contreras, D. (1996) J. Neurosci. 16,

392-417.6. Steriade, M., Curr6 Dossi, R., Pare, D. & Oakson, G. (1991) Proc.

Natl. Acad. Sci. USA 88, 4396-4400.7. Steriade, M. (1995) J. Comp. Neurol. 354, 57-70.8. Llinas, R., Grace, A. A. & Yarom, Y. (1991) Proc. Natl. Acad. Sci.

USA 88, 897-901.9. Steriade, M., Curro Dossi, R. & Contreras, D. (1993) Neuro-

science 56, 1-9.10. Llinas, R., Ribary, U., Joliot, M. & Wang, X.-J. (1994) in

Temporal Coding in the Brain, eds. Buzsaki, G., Llina's, R., Singer,W., Berthoz, A. & Christen, Y. (Springer, Berlin), pp. 251-272.

11. Bendat, J. S. & Piersol, A. G. (1980) Engineering Applications ofCorrelation and Spectral Analysis New York (Wiley, New York).

12. Amzica, F. & Steriade, M. (1995) J. Neurosci. 15, 4658-4677.13. Hubbard, J. I., Llinas, R. & Quastel, D. M. J. (1969) Electrophys-

iological Analysis of Synaptic Transmission (Arnold, London).14. Mitzdorf, U. (1985) Physiol. Rev. 65, 37-100.15. Amzica, F. & Steriade, M. (1995) J. Neurophysiol. 73, 20-39.16. Steriade, M. & McCarley, R. W. (1990) Brainstem Control of

Wakefiulness and Sleep (Plenum, New York).17. Steriade, M., Par6, D., Parent, A. & Smith, Y. (1988) Neuro-

science 25, 47-67.18. Ylinen, A., Bragin, A., Nadasdy, Z., Jando, G., Szabo, I., Sik, A.

& Buzsaki, G. (1995) J. Neurosci. 15, 30-46.19. Contreras, D. & Steriade, M. (1995) J. Neurosci. 15, 604-622.20. Jones, E. G. (1985) The Thalamus (Plenum, New York).21. Nuniez, A., Amzica, F. & Steriade, M. (1992) Neuroscience 51,

7-10.22. Curr6 Dossi, R., Pare, D. & Steriade, M. (1991) J. Neurophysiol.

65, 393-406.23. McCormick, D. A. & Prince, D. A. (1987) J. Physiol. (London)

392, 147-165.24. Steriade, M., Contreras, D., Amzica, F. & Timofeev, I. (1996) J.

Neurosci. 16, in press.

Proc. Natl. Acad. Sci. USA 93 (1996)

Dow

nloa

ded

by g

uest

on

Mar

ch 5

, 202

1