Rhythmic neuronal interactions and synchronization in the rat dorsal column nuclei

RHYTHMIC NEURONAL INTERACTIONS AND SYNCHRONIZATION IN THE RAT

DORSAL COLUMN NUCLEI

A. NUNÄ EZ,*² F. PANETSOS³ and C. AVENDANÄ O*

*Department of Morphology, School of Medicine, Universidad Autonoma de Madrid, Madrid, 28029, Spain

³Department of Applied Mathematics, School of Biology, Universidad Complutense, Madrid, 28040, Spain

AbstractÐSingle-unit and multiunit activities were recorded from dorsal column nuclei of anesthetized rats in order to study thecharacteristics of the oscillatory activity expressed by these cells and their neuronal interactions. On the basis of their ®ring ratecharacteristics in spontaneous conditions, two types of dorsal column nuclei cell have been identi®ed. Low-frequency cells (74%)were silent or displayed a low ®ring rate (1.9^ 0.48 spikes/s), and were identi®ed as thalamic-projecting neurons because they wereactivated antidromically by medial lemniscus stimulation. High-frequency cells (26%) were characterized by higher discharge rates(27.2^ 5.1 spikes/s). None of them was antidromically activated by medial lemniscus stimulation. Low-frequency neurons showeda non-rhythmic discharge pattern spontaneously which became rhythmic under sensory stimulation of their receptive ®elds (48% ofcases; 4.8^ 0.23 Hz). All high-frequency neurons showed a rhythmic discharge pattern at 13.8^ 0.68 Hz either spontaneously orduring sensory stimulation of their receptive ®elds. The shift predictor analysis indicated that oscillatory activity is not phase-locked to the stimulus onset in either type of cell, although the stimulus can reset the phase of the rhythmic activity of high-frequency cells. Cross-correlograms between pairs of low-frequency neurons typically revealed synchronized rhythmic activitywhen the overlapping receptive ®elds were stimulated. Rhythmic synchronization of high-frequency discharges was rarelyobserved spontaneously or under sensory stimulation. High-frequency neuronal ®ring could be correlated with the low-frequencyneuronal activity or more often with the multiunit activity during sensory stimulation. Moreover, the presence of oscillatory activitymodulated the sensory responses of dorsal column nuclei cells, favoring their responses.

These ®ndings indicate that thalamic-projecting and non-projecting neurons in dorsal column nuclei exhibited distinct oscillatorycharacteristics. However, both types of neuron may be entrained into an oscillatory rhythmic pattern when their overlappingreceptive ®elds are stimulated, suggesting that in those conditions the dorsal column nuclei generate a populational oscillatoryoutput to the somatosensory thalamus which could modulate and amplify the effectiveness of the somatosensory transmission.q 2000 IBRO. Published by Elsevier Science Ltd. All rights reserved.

Key words: oscillatory activity, synchronization, somatosensory system, rat.

Most of the cortical and subcortical structures of the CNSdisplay oscillatory activities in different behavioral condi-tions, covering a broad range of frequencies (for review seeRefs 34, 36±38). Neuronal oscillatory activities in the CNSmay be generated by the combination of pacemaker mechan-isms in individual cells and reciprocal synaptic interactionsamong them or within networks of cells that do not oscillateintrinsically but resonate at particular frequencies. It has beensuggested that such oscillatory activity may contribute tospike synchronization of spatially distributed neurons,which would enhance the spatial summation of the synapticpotentials that these neurons evoke in their target cells.34,35,38

Moreover, rhythmic synchronization may induce synapticplasticity because modulatory processes at the synapsesdepend critically on the relative timing of pre- and postsyn-aptic activation.12,18

The contribution of oscillations to sensory informationprocessing has been suggested at all levels of the somato-sensory system in anesthetized or awake behavinganimals.1,19,20,26,33 For instance, in the rat trigeminal system,synchronous oscillatory activity at 7±12 Hz appeared in the

brainstem, thalamus and cortex during attentive immobility,and preceded the onset of rhythmic whisker twitching usedfor tactile exploratory movements.21 Thus, synchronous activ-ity in the trigeminal system may not only encode sensoryinformation but also set the stage to optimize somatosensoryintegration.

The dorsal column nuclei (DCN) are the ®rst relay stationin the somatosensory system for the fast and precise transmis-sion to the thalamus of information about touch and positionof the trunk and limbs. Rhythmic activities in the DCN at low(,1 Hz), d (1±4 Hz) and higher frequency ranges (.4 Hz)have been described in thalamic-projecting cells and pre-sumed interneurons of anesthetized cats and rats.1,7,16,17,26

Low-frequency oscillations appear to originate in the cortex,17,39

whereas higher frequency oscillations seem to arise within theDCN.7,26 This hypothesis is in agreement with recent in vitroexperiments demonstrating that DCN cells have the capacity toproduce rhythmic activity in the absence of both cortical andperipheral inputs.23 Panetsos et al.26 demonstrated that DCNneurons express coherent oscillatory activity in the 4±22 Hzfrequency range at single unit, multiunit and local ®eld potentiallevels. These oscillations appear spontaneously in presumedlocal interneurons or during natural sensory stimulation oftheir receptive ®elds (RFs) in thalamic-projecting DCNneurons, which suggests their implication in the processing ofthe somatosensory information and in its transfer to higherlevels of the CNS. This is supported by the ®nding of rhythmicexcitatory postsynaptic potentials (EPSPs) in the ventro-posterior nuclei of the thalamus.24,31

Oscillatory activity in the dorsal column nuclei 599

599

Neuroscience Vol. 100, No. 3, pp. 599±609, 2000q 2000 IBRO. Published by Elsevier Science Ltd

Printed in Great Britain. All rights reserved0306-4522/00 $20.00+0.00PII: S0306-4522(00)00305-5

Pergamon

www.elsevier.com/locate/neuroscience

²To whom correspondence should be addressed. Tel.: 134-91-397-5319;fax: 134-91-397-5353.

E-mail address: [email protected] (A. NunÄez).Abbreviations: ACH, autocorrelation histogram; CCH, cross-correlation

histogram; DCN, dorsal column nuclei; EPSP, excitatory postsynapticpotential; HF, high-frequency; LF, low-frequency; PSTH, peristimulustime histogram; RF, receptive ®eld; S-ACH, autocorrelation histogramfollowing the subtraction of the shift predictor.

The purpose of the present study was to investigate theoscillatory activity of DCN cells and their interactions spon-taneously and during sensory stimulation, using extracellularrecordings and natural sensory stimulation. Some of the datawere presented previously in abstract form.27

EXPERIMENTAL PROCEDURES

Data were obtained from 62 urethane-anaesthetized (1.6 g/kg i.p.),10 sodium pentobarbital-anaesthetized (35 mg/kg i.p.) and ®ve keta-mine hydrochloride/xylazine-anaesthetized (100 mg and 20 mg/kg i.p.,respectively) young adult Wistar rats (from Iffa±Credo, France) ofeither sex, weighing 180±250 g. Animals were placed in a stereotaxicdevice, with control of the end-tidal CO2 concentration. The bodytemperature was maintained at 378C. To record the electro-encephalogram a macroelectrode (120 mm diameter bluntly cut insu-lated nichrome wire) was lowered 1.5 mm from the cortical surfaceinto the frontal lobe. The electroencephalogram was ®ltered between0.3 and 30 Hz and continuously monitored in the oscilloscope. Supple-mental doses of the anesthetic were given when a decrease in theamplitude of the electroencephalogram slow waves was observed.Experiments were carried out in accordance with the EuropeanCommunities Council Directive (86/609/EEC), and all efforts weremade to minimize animal suffering and the number of animals used.

Recording

The cisterna magna was opened to introduce the recording micro-electrodes at a 608 angle over the surface of the nucleus. Single unit andmultiunit recordings were performed in the DCN by means of tungstenmicroelectrodes (2±5 MV; World Precision Instruments, Sarasota, FI,USA). Microelectrodes were aimed at the gracile nuclei (A: 213.6 to214.6, L: 0.2 to 1.0 from bregma; H: 0.0 to 0.5 mm from the surface ofthe nucleus; according to the atlas of Paxinos and Watson28) or cuneatenuclei (A: 213.5 to 214.5, L: 1.5 to 2.5 from bregma; H: 0.0 to0.5 mm from the surface of the nucleus). The position of the recordingswas visually controlled under a dissecting microscope. The spikeamplitude and shape were continuously monitored on-line in an analogoscilloscope. Extracellular recordings were ®ltered (0.3±3 kHz),ampli®ed and fed to a Macintosh computer (10 kHz sample frequency)for off-line analysis. In most cases the extracellular recordingdisplayed a single unit with a spike of largest amplitude, and multiunitactivity of the neuronal population around the microelectrode, whichappeared as spikes of different and smaller amplitudes. The unit activ-ity was extracted from the extracellular records using a window dis-criminator, and the multiunit activity was selected using a lowervoltage threshold and subtracting the activity of the isolated unit.We used this procedure to compare the activity of single units withthat from the neighboring neuronal population. In many cases two unitswith distinct and stable amplitudes were recorded simultaneously.Both neurons were extracted using a window discriminator. Pairs oftungsten microelectrodes (125 mm distance; stereotrode, World Preci-sion Instruments) or home-made pairs of tungsten microelectrodes(800 mm distance) were also used in order to study synchronized activ-ity at different distances along the rostrocaudal direction.

To identify DCN cells projecting to the thalamus, bipolar stimulat-ing electrodes (120 mm diameter blunt cut nichrome wire) were aimedat the medial lemniscus (A: 2 6.5, L: 0.5±1.5, H: 8±9 mm). Anti-dromic ®ring was evoked by means of brief rectangular pulses (0.1±0.3 ms, 50±500 mA).

Somatosensory stimulation

Once single neurons had been isolated, their cutaneous RFs weremapped by a small hand-held brush while listening to their ampli®edneuronal discharges over an acoustic speaker. RFs were de®ned by thelimits at which those stimuli elicited changes in the unit dischargepattern. Precise cutaneous stimulation was performed by an electro-nically gated, short-lasting air jet (1±2 kg/cm2, of 10±300 ms dura-tion) delivered at 0.5 Hz through a polyethylene tube of 1 mm of innerdiameter. On some occasions, air-jet stimuli were triggered using thesignal of a previous spontaneous spike as a reference. Air-jet stimuliwere used as a means to obtain secure punctate and phasic naturalstimulation. The skin area eliciting the greatest response was de®nedas the functional center of the RF.

Data analysis

Statistical analyses of the recorded signals were performed off-lineusing Spike 2 software (Cambridge Electronic Design) running on aMacintosh computer. Single unit recordings were accepted for statis-tical analysis when the amplitude was three times larger than theamplitude of the multiunit activity recorded simultaneously, and the¯uctuation of the unit amplitude was lower than 10% throughout theexperiment. Summed peristimulus time histograms (PSTHs) werecalculated, using 2 ms binwidths.

Autocorrelation histograms (ACHs) were generated to quantifyrhythmicities (5 ms binwidths). ACHs had to contain three or morepeaks at regular temporal intervals to be considered as oscillating.Cross-correlation histograms (CCHs) were also used to quantify rela-tionships between DCN units as well as between unit and multiunitactivities (5 ms binwidths). To distinguish correlations due to neuronalinteractions from stimulus-induced covariations of ®ring rates, wecomputed control correlograms between responses evoked by succes-sive presentations of the stimulus (shift predictor).22,29 The ACHsincluded in the ®gures were generated following the subtraction ofthe shift predictor, in order to remove the effect of the stimulus onseton the evoked oscillatory activity (S-ACHs). All data are shown asmean^ S.E. A Wilcoxon test was used for comparisons. Statisticalsigni®cance was considered as P , 0.05.

RESULTS

General properties of dorsal column nuclei cells inspontaneous conditions

Results were obtained from 199 single unit and 77 multi-unit recordings under urethane anesthesia, 32 units and 24multiunit recordings from rats anaesthetized with pentobarbitaland 18 units and 10 multiunit recordings from rats anaesthe-tized with ketamine/xylazine. No signi®cant differences wereobserved in the spontaneous activity under different anes-thetics. DCN neurons could be grouped into two classesaccording to their ®ring rate in spontaneous conditions:low- (LF) and high-frequency (HF) cells, as previouslydescribed.25

LF neurons were more commonly found (n� 148, 74%).These cells were silent or showed very low discharge rates atrest (1.9^ 0.48 spikes/s; range 0±10 spikes/s; see Fig. 1A).These neurons were identi®ed as projecting neurons becausemost of them (26 out of 34 cells) were antidromically acti-vated from electrical stimulation of the medial lemniscus.Spikes were identi®ed as antidromic when they displayed aconstant latency after the stimulus, were blocked by a spon-taneous spike close to the stimulus, and followed high(.100 Hz) stimulation rates (Fig. 1C1±3).

In contrast, HF neurons (n� 51, 26%) were characterizedby discharge rates higher than 10 spikes/s in spontaneousconditions (27.2^ 5.1 spikes/s; range 10±80 spikes/s; Fig.1B), which also increased during sensory stimulation oftheir RFs, and by a prominent oscillatory activity (seebelow). None of these neurons was antidromically activatedfrom the medial lemniscus (n� 27). The proportion of LF andHF cells was about the same in gracile and cuneate nuclei, anddifferences in their anatomical distribution within the DCNwere not observed. Therefore, we decided to pool together theresults obtained in both nuclei as DCN neurons.

Oscillatory activity

A great majority of LF neurons (142 out of 148 cells; 96%)displayed a non-rhythmic activity in spontaneous conditions,as indicated by a ¯at ACH (Fig. 2A1), while the remaining sixcases showed rhythmic activity at 5±7 Hz. Sensory stimula-tion of their RFs (Fig. 2A3) increased the ®ring rate and

A. NunÄez et al.600

elicited a rhythmic activity in many of the LF neuronsrecorded under urethane anaesthesia (n� 71, 48%; Fig.2A2). The evoked LF neuronal oscillation had a meanfrequency of 4.8^ 0.23 Hz (range: 2.3±8.1 Hz; Fig. 3, LF).

Although brief stimulation trials elicited LF oscillations, thesewere more ef®ciently evoked by applying longer-lastingstimuli, typically repeating air-jets or by gently stroking onthe RFs with a brush. The persistence of modulation in the

Oscillatory activity in the dorsal column nuclei 601

Fig. 2. Firing pattern of a LF unit (A) and of DCN multiunit activity (B). (A) ACHs during spontaneous activity (1) and during sensory stimulation with air-jets(2). The ACH is ¯at in spontaneous conditions and becomes periodic during sensory stimulation. The PSTH of 30 responses is shown in (3). The ACH,following subtraction of the shift predictor (S-ACH), indicates that the oscillatory activity of this neuron is not phase-locked to the stimulus onset (4). (B)

Oscillatory activity is also observed at multiunit level. ACHs are ¯at at rest (1) and become periodic during sensory stimulation (2).

Fig. 1. DCN cellular types. (A, B) Raw data of representative LF and HF neurons, respectively, during spontaneous conditions. Note the rhythmicity andhigher spike ®ring of the HF neuron. (C) Antidromic activation of a representative LF neuron from medial lemniscus stimulation. Antidromic spikes showconstant latency (1; three sweeps are superimposed); they are blocked by a previous spontaneous spike (2) and follow high stimulation frequency (250 Hz; 3).

Arrowheads indicate stimulus artifact.

ACH following subtraction of the shift predictor (S-ACH; seeExperimental Procedures) indicated that the oscillatory activ-ity was not phase-locked to the stimulus onset (Fig. 2A4).

The ®ring pattern in multiunit recordings was similar tothat of LF units, because of the higher proportion of LFneurons in DCN compared to that of HF neurons (74% vs26%). Thus, ACHs of multiunit recordings were typically¯at in spontaneous conditions and revealed rhythmic peaksat 2±9 Hz during stimulation of their RFs (42 out of 77recordings, 55%; Fig. 2B1 and B2, respectively). Multiunitrecordings of neurons with small RFs on the ®ngers displayedoscillatory activity under sensory stimulation in 26 out of 42recordings (62%) and only in 16 out of 35 recordings withlarger RFs (46%).

HF neurons differed from LF neurons and from the multi-unit activity by exhibiting higher spontaneous ®ring rates (seeabove). In addition, all HF cells (n� 51; 100%) displayedrhythmic activity in spontaneous conditions, as illustratedby the raw data (Fig. 1B). The mean rate was 13.8 ^0.68 Hz at rest (7.8±27 Hz frequency range; Fig. 3). TheHF neuronal oscillation was markedly periodic in spon-taneous conditions, as shown by the very constant rhythmicpeaks in the ACH (Fig. 4A1). During sensory stimulation HFneuronal oscillations remained unchanged (Fig. 4A2). Thepersistence of the modulation in S-ACH indicated that oscil-latory activity was not triggered by stimulus onset (Fig. 4A4).This was further supported in the majority of HF neurons (36out of 51; 71%) by the absence of rhythmicity in the PSTHs(Fig. 4A3). However, some HF neurons (15 out of 51; 29%)

revealed rhythmic peaks in their PSTHs during sensory stimu-lation of their RFs, although the modulation in ACH persistedfollowing subtraction of the shift predictor (Fig. 4B1 and B2,respectively). These results suggested that the oscillatoryactivity was independent of the stimulus onset, although thestimulus could reset the phase of the rhythmic activity ofthese neurons (see also below).

Both LF (n� 12) and HF (n� 6) units, as well as multiunitrecordings (n� 10), obtained under ketamine/xylazineanesthesia also displayed oscillations similar to thoseobtained under urethane. However, under pentobarbitalanesthesia, RF stimulation evoked spike trains and increased®ring rates in LF neurons (n� 17), HF cells (n� 15) or multi-unit (n� 24) recordings, but they did not show oscillatoryactivity either in spontaneous conditions or during sensorystimulation of their RFs (data not shown).

Although it was not studied systematically, the proportionof oscillatory neurons was different according to the size oftheir RFs. RFs on the digits or toes were smaller than RFs onthe limbs or trunk. LF neurons with smaller RFs showedoscillatory activity under sensory stimulation in 29 out of54 cells (54%), while LF neurons with larger RFs displayedoscillations under sensory stimulation only in 42 out of 94cells (45%). In HF neurons the characteristics of their oscil-latory activity were similar in HF neurons independently ofRF extension, although the majority of HF neurons displayedlarger RFs (34 out of 39; 87%) which were placed proximalon the limbs or on the trunk.

Synchronous activity

Pairs of neurons were recorded with the same microelec-trode or using two tungsten microelectrodes (see ExperimentalProcedures) in order to study the temporal correlation of DCNdischarges. Neuronal pairs were selected for this study whenboth neurons displayed overlapping RFs. In spontaneousconditions, recordings of pairs of LF neurons with the samemicroelectrode (n� 35) revealed that most of them were notcorrelated or displayed a wide peak at the zero reference,indicating that both neurons tended to ®re synchronously(Fig. 5A1). During sensory stimulation CCHs displayed peri-odic peaks with a larger one at zero, indicating that bothneurons ®red rhythmically and in phase (Fig. 5A2). Thissynchronous rhythmic activity was observed in all (n� 14)LF neuronal pairs recorded from the same microelectrode thatdisplayed oscillatory activity during sensory stimulation oftheir RFs. The existence of matching RFs did not by itselfguarantee synchrony, because neurons recorded up to 800 mmaway could display a common RF, but the probability of®nding rhythmic cross-correlation features between theseneurons decreased with the distance between the two recordedLF neurons (four out of eight at 125 mm and two out of 10 at800 mm; Fig. 5C).

Synchronous rhythmic activity was never observed in pairsof LF neurons that were recorded with the same microelec-trode (n� 21) or with distant microelectrodes (n� 32), whenone cell of the LF neuronal pair did not display oscillatoryactivity under sensory stimulation of its RF. In these cases, theCCHs only revealed a peak at zero reference, indicating thatboth neurons tended to ®re synchronously during sensorystimulation. Consistently, shift predictors indicated thatmost of this synchronization was due to the common periph-eral sensory input (data not shown).

A. NunÄez et al.602

Fig. 3. Plots of the oscillatory frequency observed in LF and HF cells duringsomatosensory stimulation. LF neurons display rhythmic ®ring between 2

and 9 Hz, while HF cells display rhythmic ®ring between 7 and 26 Hz.

The activity of single LF units was also compared with theactivity of the neighboring neuronal population, recordedsimultaneously as multiunit activity. When LF neurons andthe neuronal population displayed rhythmic activity undersensory stimulation of their overlapping RFs, the CCHsdisplayed synchronous rhythmic activity in 15 out of 25cases studied (60%).

Pairs of HF neurons were seldom recorded with the samemicroelectrode (n� 10). Although in spontaneous conditionsHF cells showed an evident rhythmicity, none of the HF pairswas correlated, as indicated by ¯at CCHs (Fig. 5B1). Whenthe overlapping RFs of the HF neuronal pairs were stimulatedtheir CCHs displayed a wide peak at zero reference, thatindicated synchronous ®ring, mostly triggered by the stimulus(Fig. 5B2) and only one out of 10 recorded HF pairs displayedsynchronization (data not shown). None of the HF pairsrecorded with distant microelectrodes displayed rhythmicphase-locking (at 125 mm, n� 6 or at 800 mm, n� 11; Fig.5C).

Pairs of LF±HF neurons (n� 20) recorded with the samemicroelectrode had no correlated ®ring in spontaneous condi-tions, but a few of them (four out of 20 cases; 20%) revealedrhythmic synchronization during sensory stimulation. Theprobability of ®nding a rhythmic correlation was also lowfor neurons recorded at 125 mm apart (two out of ninecases; 22%) and was nil at 800 mm (zero out of eight cases;Fig. 5C).

HF neuronal ®ring was more often correlated with the

multiunit activity during sensory stimulation. As indicatedabove, in spontaneous conditions HF cells displayed a rhyth-mic ®ring pattern, while the multiunit activity recorded simul-taneously was non-rhythmic (Fig. 6A1±2, respectively).During sensory stimulation of their overlapping RFs, the HFneuron maintained its rhythmic activity, while a slower rhyth-mic discharge pattern (2±9 Hz) appeared in the multiunitactivity (Fig. 6B1±2). The CCHs between the HF cell andthe neuronal population were ¯at in spontaneous conditions inall tested cells (n� 17; Fig. 6A3). However, during sensorystimulation, seven out of 17 cases (41%) displayed periodicpeaks in the CCHs between the HF cell and the multiunitactivity (Fig. 6B3), indicating a correlation between the activ-ity of the HF cell and the populational DCN activity (mostlyLF neurons). The CCH displayed in Fig. 6B3 reveals a centralpeak at zero reference, and periodic peaks on the right side ofthe histogram at the same frequency as the oscillatory activityof the HF cell (Fig. 6B1). These peaks do not decay mono-tonically; instead, there are peaks of larger amplitude, those inphase with the background rhythmic multiunit activity, alter-nating with other peaks of lower amplitude, correspondingspeci®cally to the rhythmic activity of the HF neuron (Fig.6B3). This suggests that the DCN neuronal population couldmodify the rhythmic activity of HF neurons by resetting thephase of the HF oscillation and increasing the number ofspikes ®red in each HF oscillatory cycle that coincide withthe DCN population oscillation peaks.

To further demonstrate that the rhythmic activity of HF

Oscillatory activity in the dorsal column nuclei 603

Fig. 4. Oscillatory activity of HF cells. (A) A typical HF cell displays rhythmic ®ring spontaneously (1) or during sensory stimulation of its RF with air-jets (2),as shown by the ACHs. The PSTH of 30 responses is shown in (3). Rhythmic peaks in the S-ACH indicates that the oscillatory activity is not phase-locked tothe stimulus onset (4). (B) However, some HF cells display rhythmic peaks in the PSTH after the stimulus occurrence (1), indicating that the stimulus mayreset the phase of the HF oscillatory activity. Nevertheless, the persistence of the oscillatory modulation in the S-ACH suggests that most of the HF neuronal

oscillation is independent of the stimulus (2).

neurons may be modulated by the activity of LF thalamic-projecting neurons, we studied the effect of medial lemniscusstimulation on the rhythmic activity of HF cells (n� 27).Medial lemniscus stimulation did not activate these cells anti-dromically (see above), but could induce a reset of the phaseof rhythmic HF neuronal activity (12 out of 27 cells), prob-ably through antidromically, activated LF projecting neurons(Fig. 6C).

Interaction between oscillatory activity and sensoryresponses in the dorsal column nuclei

To test whether rhythmic activity in DCN neurons modu-lated their response to sensory stimulation, the ef®cacy of thestimulus (mean evoked spikes per stimulus) was calculatedfor different delays between a peak of the oscillation and thestimulus occurrence. For this purpose, air-jet stimuli weretriggered by spikes at different delays during an oscillatoryperiod.

LF neurons with sensory evoked oscillatory activitydisplayed a similar ef®cacy when the stimulus was deliveredat delays between 0 and 100 ms but increased when the delaywas longer, nearer to the period of the LF rhythmic activity.A representative case in which the somatosensory stimuluswas triggered by a previous spike at 10, 50, 100 and 250 ms

delays is shown in Fig. 7A. PSTHs displayed similarresponses in all cases except for the 250 ms delay in whichthe response was higher. This delay was close to the periodof rhythmic activity shown by this neuron, as indicatedby the ACH (Fig. 7A, inset). The mean ef®cacy of the stimu-lus (evoked spikes per stimulus) was calculated at differentdelays and results are plotted in the Fig. 8A. LF neurons withoscillatory activity (n� 6) displayed similar ef®cacy of thestimulus at delays shorter that 100 ms, but increased at longerintervals, reaching statistical signi®cance at 200±300 msdelays (Fig. 8A). These longer intervals corresponded tothe mean period (about 200 ms) of the rhythmic activitydisplayed by LF neurons (4.8 Hz, see above). However, LFneurons that did not ®re rhythmically during sensory stimula-tion of their RFs, did not modify the stimulus ef®cacyat different delays between a previous spike and the stimu-lus onset, as observed in PSTHs (Fig. 7B) or in the plot ofthe mean stimulus ef®cacy calculated in six non-oscillatingLF cells (Fig. 8B). These results suggest that sensoryresponses of LF neurons are modulated by the oscillatoryactivity.

HF cells showed different responses depending on theinterval between the stimulus onset and a previous spike ofthe neuron. The PSTHs of a typical HF cell which showed alower sensory response when the stimuli were delivered

A. NunÄez et al.604

Fig. 5. Synchronous activity between DCN cells. (A) CCHs between two neighboring LF cells display typically a single peak spontaneously (1) and periodicpeaks during sensory stimulation of their overlapping RFs (2), indicating rhythmic phase-locking between both cells. (B) CCHs between two neighboring HFcells display rhythmic correlation neither at rest (1) nor during sensory stimulation (2). The single peak observed at zero reference indicates synchronous ®ringin both neurons which is triggered by the stimulus. (C) Plot of the percentage of synchronization between LF neuronal pairs (empty bars), LF±HF pairs(hatched bars) and HF neuronal pairs (solid bars) found when neurons were recorded with the same microelectrode (0 mm) or with two microelectrodes

separated 125 or 800 mm. Synchronization between DCN cells decreases with the distance.

shortly after a previous spike (10 or 50 ms delay) comparedwith the response of the same cell when the delay was longer(100 ms) are illustrated in Fig. 7C. Also, with longer delays,the stimuli arrived close to the period of the HF oscillatoryactivity (120 ms in this case; inset in Fig. 7C), and wouldinduce an advance of the rhythmic discharge. The plot ofstimulus ef®cacy shows statistical differences at delays longerthan 50 ms in comparison with 0±20 ms (n� 6; Fig. 8C).

DISCUSSION

In the present paper we investigated the characteristicsof the oscillatory activities and neuronal interactionsinduced in DCN neurons by somatosensory stimulation oftheir RFs, as well as the modi®cation of the sensoryresponses by those oscillatory activities. Thalamic-projectingand non-projecting neurons exhibited different dischargepatterns, both in spontaneous conditions and under stimu-lation of their RFs. Our results demonstrate that both typesof neuron may be entrained into an oscillatory rhythmicpattern when their overlapping RFs are stimulated, suggestingthat in those conditions the DCN generate a populationaloscillatory output to the somatosensory thalamus whichcould modulate and amplify the effectiveness of the somato-sensory transmission.

Dorsal column nuclei neuronal types

It is known from anatomic studies that two main types of

cell exist in the DCN. The most conspicuous are large relayneurons that receive ascending branches from the dorsalcolumns and project to the ventroposterior nucleus of thecontralateral thalamus.8±10,40,41 Inhibitory interneurons whichcontain g-aminobutyric acid and/or glycine have also beendescribed.2,5,11,13,14,32

Two types of DCN neuron, LF and HF cells, have beenidenti®ed according to their electrophysiological characteris-tics.25 At least a majority of the LF neurons projects to thethalamus, because they could be antidromically activated bystimulating the medial lemniscus at the mesodiencephalicisthmus. However, there is no direct electrophysiologicalevidence that HF neurons are local interneurons. It has beenhypothesized that HF neurons could be inhibitory inter-neurons because: (i) they were not antidromically activatedby medial lemniscus stimulation; (ii) their ®ring ratedecreased during temporary deafferentation, simultaneouslywith a decrease of inhibition in LF projecting neurons;25

and (iii) they typically displayed larger RFs than LFneurons which were located proximal in the limbs andtrunk, as occurs in presumed interneurons recorded in cats.7

The present data do not exclude the possibility that HF cellsproject to the spinal cord, brainstem or cerebellum, becausethese projections from the DCN do not course through themedial lemniscus. However, the DCN neurons giving rise tolong non-lemniscal projections are mainly activated by non-primary afferents, which arise from neurons in the spinal cordthat convey convergent impulses from various sources.6,30

This would not ®t well with the fact that both HF and LF

Oscillatory activity in the dorsal column nuclei 605

Fig. 6. Synchronous activity of HF cells and DCN neuronal population. (A) ACHs of an HF cell and the multiunit activity recorded simultaneously inspontaneous conditions (1 and 2, respectively). Note the rhythmic activity of the HF cell while the DCN neuronal population shows a non-rhythmic dischargepattern. The CCH between both neuronal activities displays a single small peak at zero reference (3), suggesting that in spontaneous conditions they tend to besynchronized. (B) ACHs of the HF cell (1), the multiunit activity (2), and the CCH between them (3), during sensory stimulation of their overlapping RFs. TheHF cell maintains the same rhythmic activity while the DCN neuronal population changes to a rhythmic ®ring pattern. Consequently, the CCH shows rhythmicphase-locking between both neuronal populations. Arrowheads show alternate peaks with increased amplitudes. (C) PSTH of an HF cell activity during medial

lemniscus stimulation (20 stimuli). After the stimulus, spikes tend to appear with a constant latency at a low rate, typical of LF cells.

neurons display similar short latencies upon cutaneousstimulation.25

Oscillatory activity

The present results extend the differences between LF andHF cells, providing evidence that they also display distinctoscillatory behaviors. While HF cells showed a markedlyperiodic oscillation in spontaneous conditions which persistedduring sensory stimulation, LF neurons showed a slowerrhythmic activity only under sensory stimulation of theirRFs. Previous results have suggested that both LF and HFoscillations are generated within the DCN, because axonalrecordings of the dorsal columns showed oscillations in theprimary afferents neither spontaneously nor under sensorystimulation of the RFs, and because oscillations remainedafter decortication of the sensorimotor cortex.26 However,the mechanisms by which the stimuli induce oscillations inthe DCN are still unknown.

The shift-predictor analysis performed in the present work,which was designed to distinguish correlations due toneuronal interactions from stimulus-induced covariations of®ring rates,29 indicates that both oscillatory activitiesobserved in LF and HF cells under sensory stimulation arenot phase-locked to the stimulus onset. It seems, therefore,that the changes in the membrane potential of DCN neurons

caused by afferent sensory input induce rhythmic ®ring in LFcells, while HF cells increase their ®ring rate without alteringtheir rhythmic discharge pattern. In agreement with this, invitro studies in rats demonstrate that some DCN neurons showrhythmic activity under depolarizations due to the activationof intrinsic properties.23 Moreover, trains of electrical stimuliapplied to the dorsal columns in order to mimic the naturalinput of sensory stimuli induce depolarizations and rhythmicactivity in DCN cells, which outlast the period of stimulation.In addition, recordings of cuneate cells in anesthetized catsdemonstrate that sensory stimulation induce a long-lastingdepolarization of the membrane potential.7 In sum, resultsfrom intracellular studies strongly suggest that the naturalsensory inputs from the periphery may induce long-lastingdepolarizations in the membrane potentials of DCN cellsand rhythmic ®ring.

Spontaneous rhythmic activities in cuneothalamic cells andpresumed interneurons have been also described in anaesthe-tized cats in the same frequency band as the oscillatory activ-ity observed here in rat LF and HF cells.1,7 Amassian andGiblin1 also observed oscillations under sensory stimulationwhich were considered to be dependent on peripheral inputs,because they disappeared following deafferentation andbecause rhythmic EPSPs could be observed in in vivo intra-cellular recordings of cuneate neurons. However, our resultssupport the alternative view that the oscillatory activity is

A. NunÄez et al.606

Fig. 7. Modi®cation of sensory responses by the DCN oscillatory activity. (A) PSTHs of a representative oscillating LF cell during somatosensory stimulationwith air-jets (15 stimuli). They are calculated according to the delay between the stimulus onset (zero reference) and the previous spike (shown by an arrow).From left to right, delays are 10, 50, 100 and 250 ms, respectively, and are indicated in the histogram. The response is higher at 250 ms, which corresponds tothe period of the oscillatory activity displayed by this LF neuron, as indicated by the ACH (inset). (B) PSTHs (15 stimuli) of a non-oscillating LF neuronindicate that sensory responses are similar at different delays. Inset shows a ¯at ACH, indicating non-rhythmic activity during sensory stimulation. (C) PSTHsof a typical HF cell during somatosensory stimulation with air-jets (30 stimuli) at different delays. Sensory responses are diminished with delays of 10 and50 ms in comparison with a delay of 100 ms. Inset shows the ACH of the HF cell. Note that the 100 ms delay is close to the mean interval of its rhythmic ®ring

rate (120 ms).

generated from synaptic interactions between rhythmicneurons within the DCN. This discrepancy may re¯ect merelyinterspecies differences. The results presented here (see alsobelow) strongly suggest that neuronal interactions betweenthalamic projecting DCN neurons may be crucial to generatean oscillatory output to the thalamus that may contribute tosomatosensory information processing and to entrain non-projecting DCN neurons. Thus, deafferentation of sensoryinputs could disorganize this DCN oscillatory neuronalnetwork, which may explain the above results observed incats. Therefore, the existence of rhythmic EPSPs in themembrane potential of DCN cells instead of indicate a rhyth-mic input from the periphery could suggest that they aregenerated within the DCN through axonal collaterals, as wepostulated.

The presence of an oscillatory modulation of the DCN ®ringrate implies that the postsynaptic neuron has a heightened

sensitivity to the precise timing of sensory synaptic inputs,because the increase of ionic conductances involved in thegeneration of the pacemaker potentials could shunt theperipheral synaptic inputs. This effect is particularly relevantfor short-lasting EPSPs, which can reach threshold onlywithin a narrow temporal window. This may occur in DCNcells, where EPSPs elicited by dorsal column stimulation arebrief.23 Therefore, the effectiveness of the sensory stimulus onDCN cells could be modulated by the oscillatory activitydisplayed by these cells, as occurs in other sensory systems(see e.g. Refs 3, 4 and 42).

Certainly, this is what happens in LF oscillating neuronsand HF neurons. LF oscillating neurons showed an increase ofthe stimulus ef®cacy at intervals corresponding to the periodof their oscillation, suggesting that the populational rhythmicactivity of DCN cells may favor sensory responses. Thesame result occurs in HF cells where sensory responses aresigni®cantly increased at intervals close to their oscillatingperiod. Thus, the postulated periodic membrane potential¯uctuations in DCN cells introduce non-linearities andtemporal ®lter properties in these neurons, which could facil-itate the sensory transmission to higher levels of the somato-sensory pathway.

Rhythmic interactions and synchronization

One of the postulated roles for the oscillatory activities inthe CNS is to contribute to spike synchronization of spatiallydistributed neurons. Rhythmic interactions among oscillatoryLF neurons are commonly observed when their overlappingRFs are stimulated. All LF neurons recorded with the samemicroelectrode displayed synchronous rhythmic ®ring.However, the proportion of rhythmic phase-locking decreasedwith the distance between DCN cells. Distant LF neuronswith overlapping RFs only displayed a wide peak at zeroreference, indicating that the stimulus evoked a synchronizedspike burst. Thus, rhythmic coupling might arise presumablyfrom phase-locking of spatially adjacent oscillatory neuronsmediated by reciprocal connections among neighboringcells. The anatomical substrate of that synchronization isthe existence in rats of axon collateral branches destined tosynapse with the same or other cells of the nucleus.41 The highlevel of synchronization between LF neurons together withthe larger proportion of LF cells in the DCN result in thelarge number of multiunit recordings that express oscillatoryactivity under sensory stimulation. Thus DCN oscillationsmay rely on intra-DCN network for their generation andsynchronization with suf®cient spatial and temporal coher-ence to be observed in multiunit recordings and local ®eldpotentials.26

The HF neurons did not display synchronized rhythmic®ring either spontaneously or during sensory stimulation,indicating either that there are no synaptic interactionsbetween this type of neuron, or that they are not powerfulenough to entrain their spike activity into an oscillatorynetwork. However, in some cases (41%), HF cells werephase-locked with the slower rhythmic activity evoked inLF neurons during sensory stimulation. This may besupported by the existence of axon collaterals of projectingneurons that contact local interneurons.15 The fact that bothtypes of cell are entrained in the rhythmic activity displayedby the thalamic-projecting neurons strongly suggests thatthe DCN transform a tonic input from the periphery into a

Oscillatory activity in the dorsal column nuclei 607

Fig. 8. Changes of stimulus ef®cacy evoked by the DCN oscillatory activity.(A±C) Plots of mean stimulus ef®cacy (evoked spikes per stimulus) vsdelay of the stimulus onset to a previous spike are shown for LF oscillatingcells (n� 6), LF non-oscillating cells (n� 6) and HF cells (n� 6), respec-tively. Sensory responses of LF oscillating cells are increased at intervalslonger than 100 ms in comparison with 0±20 ms interval, reaching statis-tical signi®cance at 200±300 ms delay (A), while no signi®cant changes areobserved in LF non-oscillating cells (B). HF cells increase their stimulusef®cacy at delays .50 ms in comparison with 0±20 ms interval. Plots

display the mean^S.E.M. (*P , 0.05).

populational oscillatory output to the somatosensory thalamusthat entrains all kinds of functionally related DCN neuronsduring sensory stimulation of a common RF. A driving in¯u-ence of LF neurons on the activity of HF neurons is alsoindicated by the reset of the rhythmic activity in the latterby the electrical stimulation of the medial lemniscus, whichevokes antidromic activation of LF but not HF neurons.

The existence of a populational oscillatory activity inDCN during sensory stimulation may induce a rhythmicsynaptic input in the somatosensory thalamus. Correspond-ingly, intracellular recordings in rat ventroposterior nucleiof the thalamus revealed rhythmic subthreshold eventswhich were identi®ed as EPSPs, probably originating inthe DCN because they disappeared after lesioning theDCN.31 They appeared spontaneously or during sensorystimulation of their RFs, at about 20 Hz (see Fig. 4 inRef. 24; see also Ref. 31). The different ®ring rate of LFneurons and of those rhythmic EPSPs described in the thala-mus by Pinault and DescheÃnes31 could be due to the use oflower doses of urethane (1.6 to 1.4 g/kg), which could inducespontaneous ®ring and faster oscillatory activities of DCNcells.

Functional consequences of oscillations in the somatosensorysystem

It has been hypothesized that oscillations in the somato-sensory cortex could facilitate an association between func-tionally related cells.19,20,33 This ªassociation hypothesisº mayalso be applied to the DCN. Sensory inputs that arrive to DCNneurons which are spatially distributed, but share commonRFs, may be integrated into an oscillatory neuronal network.The rhythmically synchronized DCN neurons will induceEPSPs within a narrow temporal window in selected thalamiccells, thus increasing their chance to generate spikes. Thus,the rhythmic synchronization may enhance the contrastbetween the evoked EPSPs from rhythmically entrainedneurons and the synaptic outputs from non-rhythmic neurons.Consequently, DCN oscillations may contribute to the ®nefocusing of sensory responses in the somatosensory system,enhancing the activity of functionally related neurons and®ltering out the irrelevant sensory inputs from the periphery.

AcknowledgementsÐThis work was supported by CICYT (SAF96-0031) and CAM (08.5/0023/98) grants.

REFERENCES

1. Amassian V. E. and Giblin D. (1974) Periodic components in steady-state activity of cuneate neurones and their possible role in sensory coding.J. Physiol., Lond. 243, 353±385.

2. Barbaresi P., Sprea®co R., Frassconi C. and Rustioni A. (1986) GABAergic neurons are present in the dorsal column nuclei but not in the ventroposteriorcomplex of the rat. Brain Res. 382, 305±326.

3. Barrio L. C. and BunÄo W. (1990) Temporal correlations in sensory±synaptic interactions: example in cray®sh stretch receptors. J. Neurophysiol. 63,1520±1528.

4. Barth D. S. and Di S. (1991) Laminar excitability cycles in neocortex. J. Neurophysiol. 65, 891±898.5. Blomqvist A. and Westman J. (1976) Interneurons and initial axon collaterals in the feline gracile nucleus demonstrated with the rapid Golgi technique.

Brain Res. 111, 407±410.6. Brown A. G., Brown P. B., Fyffe R. E. W. and Pubols L. M. (1983) Receptive ®eld organization and response properties of spinal neurones with axons

ascending the dorsal columns in the cat. J. Physiol., Lond. 337, 575±588.7. Canedo A., Martinez L. and MarinÄo J. (1998) Tonic and bursting activity in the cuneate nucleus of the chloralose-anesthetized cat. Neuroscience 84,

603±617.8. DeBiasi S., Vitellaro-Zuccarello L., Bernardi P., Valtschanoff J. G. and Weinberg R. (1994) Ultrastructural and immunocytochemical characterization of

terminals of postsynaptic ascending dorsal column ®bers in the rat cuneate nucleus. J. comp. Neurol. 353, 109±118.9. Ellis L. C. and Rustioni A. (1981) A correlative HRP, Golgi and EM study of the intrinsic organization of the feline dorsal column nuclei. J. comp.

Neurol. 197, 341±367.10. Fyffe R. E., Cheema S. S., Light A. R. and Rustioni A. (1986) Intracellular staining study of the feline cuneate nucleus. II. Thalamic projecting neurons.

J. Neurophysiol. 56, 1284±1296.11. Heino R. and Westman J. (1991) Quantitative analysis of the feline dorsal column nuclei and their GABAergic and non-GABAergic neurons. Anat.

Embryol. 184, 181±193.12. Huerta P. T. and Lisman J. E. (1995) Bidirectional synaptic plasticity induced by a single burst during cholinergic theta oscillations in CA1 in vitro.

Neuron 15, 1053±1063.13. Lue J. H., Shieh J. Y., Wen C. Y., Chen K. N. and Chan S. A. (1994) GABAergic boutons establish synaptic contacts with the soma and dendrites of

cuneothalamic relay neurons in the rat cuneate nucleus. Expl Brain Res. 98, 13±20.14. Lue J. H., Jiang S. Y., Shieh J. Y. and Wen C. Y. (1996) The synaptic interrelationships between primary afferent terminals, cuneothalamic relay neurons

and GABA-immunoreactive boutons in the rat cuneate nucleus. Neurosci. Res. 24, 363±371.15. Lue J. H., Jiang S. Y., Shieh J. Y., Ling E. A. and Wen C. Y. (1997) Multiple inputs of GABA-immunoreactive neurons in the cuneate nucleus of the rat.

Neurosci. Res. 27, 123±132.16. MarinÄo J., Martinez L. and Canedo A. (1996) Coupled slow and delta oscillations between cuneothalamic and thalamocortical neurons in the chloralose

anesthetized cat. Neurosci. Lett. 219, 107±110.17. MarinÄo J., Aguilar J. and Canedo A. (1999) Cortico-subcortical synchronization in the chloralose-anesthetized cat. Neuroscience 93, 409±411.18. Markram H., LuÈbke J., Frotscher M. and Sakmann B. (1997) Regulation of synaptic ef®cacy by coincidence of postsynaptic APs and EPSPs. Science 275,

213±215.19. Murthy V. N. and Fetz E. E. (1992) Coherent 25- to 35-Hz oscillations in the sensorimotor cortex of awake behaving monkeys. Proc. natn. Acad. Sci.

USA 89, 5670±5674.20. Murthy V. N. and Fetz E. E. (1996) Synchronization of neurons during local ®eld potential oscillations in sensorimotor cortex of awake monkeys.

J. Neurophysiol. 76, 3968±3982.21. Nicolelis M. A. L., Baccala L. A., Lin R. C. S. and Chapin J. K. (1995) Sensorimotor encoding by synchronous neural ensemble activity at multiple levels

of the somatosensory system. Science 268, 1353±1358.22. Nowak L. G., Munk M., Nelson J. I., James A. C. and Bullier J. (1995) Structural basis of cortical synchronization. 1. Three types of interhemispheric

coupling. J. Neurophysiol. 74, 2379±2400.23. NunÄez A. and BunÄo W. (1999) In vitro electrophysiological properties of rat dorsal column nuclei neurons. Eur. J. Neurosci. 11, 1865±1876.24. NunÄez A., Barrenechea C. and AvendanÄo C. (1994) Spontaneous activity and responses to sensory stimulation in ventrobasal thalamic neurons in the rat:

an in vivo intracellular recording and staining study. Somatosens. Motor Res. 11, 89±98.

A. NunÄez et al.608

25. Panetsos F., NunÄez A. and AvendanÄo C. (1997) Electrophysiological effects of temporary deafferentation on two characterized cell types in the nucleusgracilis of the rat. Eur. J. Neurosci. 9, 563±572.

26. Panetsos F., NunÄez A. and AvendanÄo C. (1998) Sensory information processing in the dorsal column nuclei by neuronal oscillators. Neuroscience 84,635±639.

27. Panetsos F., NunÄez A. and AvendanÄo C. (1999) Somatosenory stimulation induces rhythmic interactions in rat dorsal column nuclei cells. Soc. Neurosci.Abstr. 25, 403.

28. Paxinos G. and Watson C. (1986) The Rat Brain in Stereotaxic Coordinates. Academic, New York.29. Perkel D. H., Gerstein G. L. and Moore G. P. (1967) Neuronal spike trains and stochastics point processes. Simultaneous spike trains. Biophys. J. 7,

419±441.30. Pierce J. P., Weinberg R. J. and Rustioni A. (1990) Single ®ber studies of ascending input to the cuneate nucleus of cats: II. Postsynaptic afferents. J. comp.

Neurol. 300, 134±152.31. Pinault D. and DescheÃnes M. (1992) The origin of rhythmic fast subthreshold depolarizations in thalamic relay cells of rats under urethane anaesthesia.

Brain Res. 595, 295±300.32. Popratiloff A., Valtschanoff J. G., Rustioni A. and Heinberg R. J. (1996) Colocalization of GABA and glycine in the rat dorsal column nuclei. Brain Res.

706, 308±312.33. Roy S. and Alloway K. D. (1999) Synchronization of local neural networks in the somatosensory cortex: a comparison of stationary and moving stimuli.

J. Neurophysiol. 81, 999±1013.34. Singer W. (1993) Synchronization of cortical activity and its putative role in information processing and learning. A. Rev. Physiol. 55, 349±374.35. Singer W. and Gray C. M. (1995) Visual feature integration and the temporal correlation hypothesis. A. Rev. Neurosci. 18, 555±586.36. Steriade M. (1993) Cellular substrates of brain rhythms. In Electroencephalography (eds Niedemyer E. and Lopes da Silva F.). Williams & Wilkins,

Baltimore.37. Steriade M., Gloor P., Llinas R., Lopes da Silva F. H. and Mesulam M. M. (1990) Basic mechanisms of cerebral rhythmic activities. Electroenceph. clin.

Neurophysiol. 76, 481±508.38. Steriade M., McCormick D. A. and Sejnowski T. J. (1993) Thalamocortical oscillations in the sleeping and aroused brain. Science 262, 679±685.39. Steriade M., NunÄez A. and Amzica F. (1993) A novel slow (,1 Hz) oscillation of neocortical neurons in vivo-depolarizing and hyperpolarizing

components. J. Neurosci. 13, 3252±3265.40. Tan C. K. and Lieberman A. R. (1978) Identi®cation of thalamic projection cells in the rat cuneate nucleus: a light and electron microscopy study using

horseradish peroxidase. Neurosci. Lett. 10, 19±22.41. Valverde F. (1966) The pyramidal tract in rodents. A study of its relations with the posterior column nuclei, dorsolateral reticular formation of the

medulla, and cervical spinal cord (Golgi and E.M. observations). Z. Zellforsch. 71, 297±363.42. Volgushev M., Chistiakova M. and Singer W. (1998) Modi®cation of discharge patterns of neocortical neurons by induced oscillations of the membrane

potential. Neuroscience 83, 15±25.

(Accepted 15 June 2000)

Oscillatory activity in the dorsal column nuclei 609